The problem arises when criticism of political decisions becomes confused with the delegitimisation of science. In recent years, in certain sections of the centre-right, legitimate scepticism has sometimes been accompanied by an undermining of the medical and scientific communities and, in some cases, of well-established knowledge. It is a dangerous short circuit.

Politics has the task of making decisions. Science has the task of studying the facts on the basis of the available evidence. These are different spheres. A minister may make a mistake, just as a scientist or a technical committee may. But acknowledging a scientific error does not mean replacing the scientific method with a political conviction.

Science, after all, is not democratic: it does not decide by majority vote. A piece of data does not become false simply because it is unpopular, nor does evidence become true simply because it garners widespread support. Scientific knowledge progresses through data, verification, debate and the ability to correct itself when new evidence emerges. It is precisely this ability to correct errors that is one of its greatest strengths.

The parliamentary commission of inquiry into Covid should also move in this direction. Investigating the handling of the pandemic is both legitimate and necessary: we need to understand which decisions worked, which had negative consequences, what political and administrative responsibilities have emerged, and where the state has shown its weaknesses.

But its aim should be, above all, to learn from mistakes in order to be better prepared for the next emergency. The risk, however, is that a commission of inquiry may be perceived as a tool to fuel a political battle against the scientific community, taking on a more ideological than investigative stance. That would be a missed opportunity.

The question we should be asking is not just who was at fault, but what we need to change. How can we improve protocols? How can we make data usage more transparent? How can we communicate more effectively during a crisis? How can we protect public health, education, the economy and individual freedoms all at the same time?

A political force that aspires to govern should aim to address these questions without turning science into a political adversary. Catering to the most radical elements of one’s electorate may yield immediate support, but risks leaving behind a legacy of mistrust in institutions that, in the long run, harms everyone.

Criticising the lockdown does not mean being anti-science. Challenging a public health measure does not mean denying reality. Calling for transparency regarding the data does not mean undermining the authority of doctors. But turning mistrust of science into a political identity is another matter entirely.

The lessons learnt from the Covid crisis should serve to build a country that is better prepared, not simply more divided. Politicians have the right – and the duty – to assess the decisions that have been taken. But they must do so without confusing criticism of those in power with a denial of the facts. Because it is one thing to hold people to account for the mistakes made during a pandemic. It is quite another to turn science into an enemy. The former is democracy. The latter is populism.

Ecology in recent decades has provided many reasons to be wary of this portrayal. Ecosystems have a history, and what happens during a disturbance can alter the conditions on which all subsequent developments will depend. A species that has gone extinct may have left behind a different community; a different community may have transformed the soil or water; those transformations may continue to exert their effects even after the cause that set the process in motion has been removed. Returning to the starting point therefore requires something other than simply retracing one’s steps, and in some cases that previous path may no longer be available.

Few places illustrate this problem better than Yellowstone, which for years has been used to tell precisely the opposite story.

At the beginning of the 20th century, wolves were systematically eradicated from the national park as part of predator control policies that were considered normal at the time. The last known pack was culled in the 1920s, and for around seventy years the wolf remained virtually absent from Yellowstone. Meanwhile, the elk – the large North American deer that is one of the wolf’s main prey species – exerted increasing pressure on the vegetation, particularly on willows and other plants in riparian areas.

When, in 1995, the United States decided to reintroduce wolves by transferring them from Canada, one of the most successful ecological stories of recent decades began to take shape. The wolves preyed on elk and reduced their numbers; their presence also altered the behaviour of their prey, which would frequent areas where the risk of being killed was greater less frequently; the willows were thus able to regrow, the beavers found food and material for their dams, the dams altered the course of the waterways once more, and the recovery of vegetation reduced bank erosion. By the end of this process, the return of a predator had even altered the rivers.

The scientific basis for this story was sound. In the early years following the reintroduction, several researchers investigated the possibility that Yellowstone was exhibiting a trophic cascade – that is, the propagation of effects exerted by a predator through the lower levels of the food web. Other studies focused on the ‘ecology of fear’, according to which the presence of the predator might have altered the way prey used the space even before reducing their numbers. These were reasonable ecological hypotheses, subject to normal scientific debate and difficult to distinguish from the other processes affecting the park at the same time.

The transformation took place when this complexity was condensed into a linear narrative.

In 2013, George Monbiot took Yellowstone to the TEDGlobal stage during a talk dedicated to rewilding – the restoration of ecological processes through the reintroduction of extinct species and the reduction of human intervention. The case of the wolves featured prominently in the talk: after seventy years of absence, the return of this predator would have reduced the deer population, encouraged the regrowth of trees along watercourses and allowed beavers to return, ultimately leading to changes in the very shape of the rivers.

The following year, that section of the speech was turned into the video "How Wolves Change Rivers", featuring spectacular footage of Yellowstone, Monbiot’s voiceover and a narrative flow so straightforward that it could be followed without any knowledge of ecology. The video reached tens of millions of people, and its central argument became part of environmental outreach, educational materials and popular conservation culture.

Its strength lay in the perfect symmetry of the narrative. Man had removed the wolf and Yellowstone had deteriorated; man had reintroduced the wolf and Yellowstone had begun to recover. What had been taken away was restored, and the system resumed its path towards its previous state. The removal and reintroduction appeared as two inverse operations, separated by seventy years and linked by the same causal chain, traversed first in one direction and then in the other.

This idea coincided perfectly with a much more general belief: that nature possesses a state of its own, from which we can cause it to deviate, but to which it would be able to return as soon as we restore what we have taken away. The ability of ecosystems to reorganise themselves was thus interpreted as a capacity for self-repair, attributing to nature a sort of memory of its previous state.

Yellowstone seemed to provide the evidence that was missing to support this belief.

Data collected in subsequent years painted a different picture. Wolves certainly preyed on elk and contributed to the decline in their population, alongside other large carnivores and human hunting outside the park’s boundaries. The behavioural effects attributed to fear of predators proved to be variable and context-dependent, as a herbivore must continue to forage and chooses where to do so based on the distribution of resources as well as the risk of being killed. Vegetation showed recovery in some areas and much more modest responses in others.

Daniel Stahler, who heads the Yellowstone wolf programme, described these effects using a phrase far less suited to a viral video: rather than a major ‘trophic cascade’, in many parts of the system a ‘trophic trickle’ is observed.

The most significant finding comes from an experiment that began in 2001 and continued for twenty years, published in Ecological Monographs by N. Thompson Hobbs and colleagues. The authors studied willow communities in the northern part of Yellowstone to understand why the return of large carnivores had not restored the riparian environment that existed prior to their elimination.

To understand this, one needs to look at what happened during the decades when the wolves were absent.

The willows in riparian areas have been linked to beaver activity through a relationship that has endured over time. Beavers used the willows both as food and as building material; their dams slowed the flow of water and raised the water table locally, creating conditions favourable to the growth of the willows themselves. The vegetation therefore supported the beavers’ activity, which in turn helped to maintain the environment necessary for the vegetation.

The pressure exerted by the elk kept the willows short for a long time and reduced the resources available to beavers. As the number of dams decreased, the watercourses also changed: river channels became deeper, the water table fell in several areas, and conditions favourable to the growth of riparian plants diminished. A process that began with a change in the food web had thus gradually transformed the physical environment as well.

When the wolves returned to Yellowstone, they found the result of those seventy years.

Hobbs and colleagues experimentally separated the two main factors. In some plots, they prevented large herbivores from reaching the willows by erecting fences; in others, they built artificial dams that mimicked the hydrological effects produced by beavers. The design allowed for a comparison between normal conditions and those in which grazing was reduced, and each situation could be studied with or without the restoration of the hydrological conditions associated with the dams.

The question was crucial. If the problem had essentially stemmed from pressure from the elk, reducing grazing should have allowed the willows to return to their previous condition. This was precisely the outcome that the reintroduction of large carnivores was intended to promote.

After twenty years of experimentation, simply excluding herbivores improved the growth of the willows without restoring the historic community. Altering the hydrological conditions also had a significant effect. The willows reached the heights characteristic of the environment being restored when the two measures were combined: reducing grazing whilst simultaneously restoring the conditions previously created by beaver dams.

The return of the predators had not erased what had happened during their absence, because in the meantime the system had acquired new constraints.

In ecology, this phenomenon is described by the concept of hysteresis. When a system exhibits hysteresis, the trajectory it follows after a disturbance depends on its previous history, because certain variables have changed during the transition and continue to influence what may happen subsequently. Returning the factor that caused the change to its initial value does not, therefore, force the system to retrace its steps.

The model proposed by Hobbs and colleagues illustrates the difference well. In a very simple food chain, consisting of a predator, a herbivore and a plant, one can imagine that the removal of the predator would increase the number of herbivores and reduce the number of plants, whilst the return of the predator could rapidly reverse that sequence. Real-world ecosystems comprise organisms that alter the environment in which they live and processes that operate over different timescales. During the absence of predators, herbivores can therefore contribute to changes in the habitat that persist even after the predators return.

Yellowstone provides a particularly clear example of why the beaver is an ‘ecosystem engineer’: through its dams, it directly alters the hydrology. When its activity declines, this changes a variable that will continue to influence the vegetation even if the pressure from the elk is subsequently reduced.

Meanwhile, the herbivore community has also changed. The elk population has declined sharply compared with the 1990s, whilst the bison population has increased and now accounts for a much larger proportion of the biomass of large herbivores. Bison are far less vulnerable to wolves and continue to exert pressure on the vegetation. Consequently, the food web into which the predator has been reintroduced is also different from the one that existed before its elimination.

Talking about hysteresis does not mean claiming that every transformation is definitively irreversible. Hobbs’s experiment shows that, by intervening in hydrology and herbivore pressure, it is possible to shift the system once again and encourage strong growth in willows. It means recognising that restoration does not consist of undoing the initial intervention, because the system we are acting upon today is the product of what has happened in the meantime.

This distinction profoundly alters the meaning of the expression ‘nature’s self-repair’. A disturbed ecosystem continues to organise itself according to the prevailing conditions and may stabilise in a configuration different from the previous one. Nature does not retain a blueprint of the lost state and does not work to restore it. It retains the consequences of the processes that have taken place.

A carved-out watercourse remains carved out even as the elk population declines; a lowered water table continues to affect the vegetation even after the wolves have returned. If beavers cannot find enough willow, they do not build the dams that would encourage the growth of the willow itself. The past thus becomes part of the causality of the present.

Even the concept of balance in nature becomes dangerous when it is used as if it were describing a general property of ecosystems. An ecological system can remain in different configurations for a long time and can develop mechanisms that cause a state resulting from a previous disturbance to persist. Stability offers no guarantee that the stabilised state will be the one we wish to preserve.

The implications for conservation are very serious. If degradation and restoration were symmetrical processes, we could regard the loss of a species or the transformation of a habitat as temporary damage, because it would be sufficient to subsequently eliminate the original cause and leave it to the system to rebuild what had been lost. Hysteresis removes this guarantee, because during degradation the ecosystem continues to change, and every change alters the conditions on which recovery will depend.

Restoration may therefore require interventions that no one would have foreseen at the start of the transformation and may depend on processes that have become difficult to reconstruct. In some systems, thresholds may be exceeded beyond which a return to the previous state requires conditions that no longer exist.

Preservation remains a top priority in this regard – one that no promise of restoration can replace.

The Yellowstone case also shows just how easily a sound scientific question can turn into a false certainty when it encounters a narrative that confirms what we already wish to believe. Trophic cascades do exist and the ecological effects of large predators are real; the oversimplification arose when a network of time-dependent relationships was reduced to a linear chain in which each step could be traversed in both directions.

The success of "How Wolves Change Rivers" also stemmed from this implicit reversibility. The story suggested that by removing one element, we had caused degradation; by reintroducing that same element, we would set the entire system in motion in the opposite direction. It was a perfect fit for our confirmation bias, because it transformed the complexity of ecology into a demonstration of nature’s innate ability to restore its own balance.

Twenty years of experimentation have taught us a different lesson. Ecosystems do not wait for us to intervene in order to continue changing, nor do they revert to their past state when we remove the cause that had transformed them. Every phase of their history shapes the conditions of the next phase.

The wolf has returned to Yellowstone. But the time that has passed during its absence has not.

And it is precisely for this reason that destroying and rebuilding are not equivalent actions, whilst preserving what still exists remains the safest form of ecological intervention possible.

A major study published in Nature now seeks to verify the extent to which this explanation corresponds to the observed distribution of emerging diseases. Rory Gibb and his colleagues have compiled 58,319 epidemic events relating to 32 infections, observed in 169 countries between 1910 and 2022. The bulk of the data relates to the last two decades, when epidemiological surveillance became more widespread. The sample includes a wide variety of infections, ranging from dengue to Ebola, from yellow fever to the plague, and including Nipah, mpox and Lyme disease. For each outbreak, the researchers reconstructed the environmental characteristics of the location where it had been recorded, seeking to understand which conditions were most frequently associated with the emergence of an outbreak. The crucial issue, however, predates any ecological interpretation: an outbreak can only appear in a database when someone is able to recognise it. This introduces a significant distortion. Where the healthcare system is more accessible, a disease is more likely to be diagnosed and therefore to be included in the data. Where reaching a healthcare facility takes a long time, the same event may remain unnoticed. The map of observed outbreaks therefore also bears the mark of observational capacity.

The authors sought to correct for this effect by introducing variables into the models that describe the accessibility of healthcare services. The result is very robust. In models where it was possible to estimate journey times, every additional hour required to reach a healthcare facility corresponded to a median 32 per cent reduction in the probability of an outbreak appearing in the data. That figure varies greatly from one disease to another and therefore does not represent a general rule. However, it shows just how profoundly the geography of an infectious disease outbreak depends on the ability to detect it.

This point changes the way in which many maps of zoonoses should be interpreted. A region in which few outbreaks are recorded may genuinely be less exposed, or it may simply be a region where surveillance is less effective. When this distinction is not taken into account, it is easy to attribute to the environment factors that depend, at least in part, on the structure of the healthcare system.

After correcting this distortion as far as possible, the authors nevertheless find an ecological signal. However, no single cause emerges that can explain all the diseases under consideration in the same way.

Landscape structure is particularly significant in certain categories of infection. Areas where natural habitats and human-altered zones frequently intersect are associated with a higher risk for a substantial proportion of the diseases studied. The biological rationale is plausible: when the organisation of an ecosystem changes, so too do the opportunities for contact between animal hosts, vectors and humans.

This pattern is particularly evident in infections transmitted by arthropods. For a disease transmitted by mosquitoes or ticks, the environment has a direct impact on the distribution of the vector and, consequently, on the likelihood of the pathogen circulating. In these systems, certain ecological characteristics tend to produce similar effects even across different diseases.

One of the recurring patterns concerns long-term changes in precipitation. In several vector-borne infections, progressively drier conditions are associated with an increased risk. The interpretation depends on the biological system under consideration, as water availability affects the ecology of the species involved in different ways.

With regard to other factors that have been much discussed in recent years, the picture appears far less uniform. Recent deforestation shows a statistically significant association in only a minority of the diseases examined. Nor do temperature variations produce a common signature stable enough to be observed across the entire set of pathogens studied.

This does not allow us to conclude that deforestation or climate change are irrelevant to the emergence of infections. It means that their effect does not appear to be a universal mechanism, recognisable in the same way across profoundly different biological systems.

The distinction becomes even more evident in zoonoses where the pathogen is transmitted directly from an animal to humans. For infections such as Ebola, MERS or mpox, there is no common combination of environmental factors comparable to that observed in many vector-borne diseases. In these cases, the risk appears to depend much more on the specific ecology of the animal reservoir and the manner in which contact with humans occurs.

The general category of zoonoses therefore encompasses biological mechanisms that are too diverse to be described by a single environmental explanation. The fact that a pathogen originates from an animal does not imply that its transmission to humans depends on the same conditions that facilitate the transmission of another pathogen.

The study also has significant limitations. To make very different datasets comparable, the authors have often reduced the information to the presence or absence of at least one case in a particular location and in a particular year. This means that the scale of the outbreak is lost, and for many rare infections it becomes difficult to accurately reconstruct the timeline between environmental change and the emergence of the disease.

The authors partially verified the robustness of this approach using more comprehensive U.S. data for four arboviral diseases. In that context, the simplified analysis captured most of the key geographical determinants that emerged when incidence rates were considered directly.

The resulting picture is more complex than the general explanations that often accompany discussions of emerging diseases. Human impact on ecosystems matters, but its effect depends on the biology of the pathogen and the way in which transmission occurs. Above all, what matters is the ability to see what is happening.

This is why surveillance plays a crucial role in preventing future epidemics. Where a disease is detected early, it becomes possible to trace its origin and take action before transmission spreads. Where surveillance is weak, our understanding of the ecology of epidemics remains distorted.

The geography of emerging diseases thus reveals two inseparable phenomena: where pathogens manage to reach us, and where we manage to detect them. Before attributing a gap on the map to the absence of a disease, we must be certain that someone in that place was able to detect it.

Just as these figures highlight how dangerous epidemiological invisibility is, the international system designed to reduce it is undergoing a period of contraction. In 2025, the WHO asked its national offices to assess the effects of the sharp reduction in international health aid. Of the 108 countries surveyed, 66 per cent had already reported disruptions to public health surveillance, whilst 70 per cent reported consequences for emergency preparedness and response. By the end of that same year, the Organisation estimated that external health aid would decline by a total of 30–40 per cent compared with 2023.

The WHO, too, now has fewer resources than were anticipated a few years ago. The United States’ decision to withdraw from the Organisation and to halt its funding has had a significant impact, alongside the reduction in development aid decided upon by other countries. In April 2025, Director-General Tedros Adhanom Ghebreyesus indicated a projected salary shortfall of between 560 and 650 million dollars for the two-year period 2026–2027; the budget was subsequently reduced from 5.3 to 4.2 billion dollars, representing a 21 per cent cut compared with the original proposal. The resulting restructuring has led to a reduction in staff numbers and in the number of programmes the Organisation can support.

The effects have precisely targeted the functions that the study published in Nature identifies as essential. In its 2025 report on findings, the WHO states that financial constraints have limited the surveillance networks and monitoring required under the International Health Regulations; in many countries, laboratory activities and the capacity for early detection of emergencies have been scaled back. In the same year, however, the Organisation’s global systems still screened around 5.4 million potentially relevant alerts and verified around 693 possible health threats with governments. These figures illustrate the scale of the apparatus required to transform the local emergence of a disease into information available to the rest of the world.

There are also investments moving in the opposite direction: in 2025, the WHO updated the EIOS (Epidemic Intelligence from Open Sources) system, which is now used by more than 110 countries to detect early signs of potential health threats. However, its own strategy identifies the availability of stable, long-term funding as one of the necessary conditions. A system capable of gathering global signals is, in fact, only useful if, at the outset, there are people and structures capable of generating those signals.

This creates a very clear contradiction between what we are learning about the origins of epidemics and the direction in which some international health policies are heading. The work by Gibb and colleagues shows that one of the major obstacles to understanding infectious risk lies in the unequal probability of an outbreak being detected. Reducing surveillance in countries with more fragile healthcare systems makes that probability even more unequal and expands the areas in which a new pathogen can circulate before being recognised.

This issue also directly affects countries with much stronger healthcare systems. An unnoticed outbreak in a distant region does not necessarily remain confined to the place where it first appeared, whilst information gathered at an early stage can travel much more quickly than the pathogen itself, enabling other countries to prepare. International surveillance therefore constitutes a form of shared protection, the value of which also depends on the weakest links in the network.

If the key epidemiological lesson to be learnt from this data is that we cannot prevent what we cannot see, then reducing the world’s ability to see today means deliberately accepting a poorer understanding of the next infectious threat.

Just a few days ago, the public debate was dominated by catastrophic scenarios following statements by researchers at Anthropic – with internal warnings that AI could lead to human extinction within the decade – which were followed by official statements from CEO Dario Amodei, who called for a controlled slowdown in the sector, and public support from figures such as Elon Musk. Yet, with disarming nonchalance, the very same premises are now being turned on their head to foreshadow the imminent eradication of almost every human disease. Amidst this swing between catastrophism and biological utopia lies the latest announcement from Anthropic, the company developing the Claude model.

The scientific fact, stripped of all propaganda, is undoubtedly intriguing. Set loose on a vast genomic database with just a single guiding prompt, Claude co-ordinated around 950 agents in parallel for 21 consecutive hours. It sifted through hundreds of thousands of DNA sequences, narrowed down the field with the tenacity of a bloodhound, and finally identified something that had eluded human biologists: an enzymatic system based on a reverse transcriptase associated with a series of uniformly spaced repetitive sequences. The researchers have named it ART (array-associated reverse transcriptases). A molecular arrangement that closely resembles the structure of CRISPR, the bacterial immune system that has revolutionised gene editing over the last decade.

In biological terms, the discovery is genuine. Even Feng Zhang, a professor at MIT and one of the pioneers of molecular scissors, acknowledged this, describing the result as “genuinely exciting”. And Amodei himself, with self-deprecating humour measured down to the millimetre, commented that the discovery of this molecular machine in bacteriophages represents “a piece of work I would have been proud of as a PhD student”. The machine read the literature, sifted through the data, spotted the anomaly and even suggested the bench-top experiments that real-life scientists then carried out in the laboratory to confirm that the array is indeed expressed in short RNA sequences.

And yet, reading the pre-print released by Anthropic with a touch of scepticism, it is difficult to separate the scientific revelation from the public relations moment. The publication comes hot on the heels of yet another round of controversy and reports on the ‘existential risks’ associated with next-generation models. Anthropic’s move has a distinctly defensive flavour: what better way to allay fears of a destructive AI than to show the world that your model is laying the foundations for the medicine of the future?

What is needed, in fact, is a touch of secular clarity. As the authors of the paper themselves admit with a modicum of modesty, the precise biological function of ART remains unknown. We do not know whether it really functions as an editing tool, nor whether it will ever have any concrete biotechnological utility or be applicable to humans. To turn the identification of a genomic pattern into the implicit promise of a universal cure is to confuse the first stage of a Grand Prix with the finish line. Biomedical progress is an gruelling marathon stretching from basic biology to clinical trials, via the complex process of delivering medicines to patients; a journey where failures in the test tube number in the thousands.

This in no way detracts from the value of the method. Seeing an agent-based architecture process 210 million tokens to carry out, in less than a day, a sifting task that would have taken a team of researchers months is confirmation that the trajectory of AI applied to the life sciences is not merely a marketing ploy. Amodei often refers to his essay "Machines of Loving Grace" and the exponential trajectory of models which, having stumbled over secondary school equations in 2023, have now reached the point where they are tackling open problems in mathematics. Biology could follow a similar trajectory.

But in the meantime, amidst press releases and pre-prints – whether the brainchild of the model itself or the PR department’s strategy – we would do well to keep a cool head. Artificial intelligence will not wipe us out by 2030, but nor will it cure all the world’s diseases within the same timeframe. In between lies real science: made up of brilliant algorithmic insights, human laboratories tasked with verifying the data and, above all, a public that must learn to distinguish between the advancement of knowledge and a sophisticated public relations campaign.

The first specimen had been observed more than ten years ago by the biologist Paola Nogales-Ascarrunz, who was contacted by staff at the Senda Verde animal shelter in the Bolivian Andes. The shelter had in fact been entrusted with the care of a ‘tiger cat’, brought in by a man who had tried unsuccessfully to tame it for a year. Initially, the researcher had mistakenly classified the feline as Leopardus tigrinus, a species native to the area, but she then noticed the irregular and ‘open’ ‘rosettes’ (markings on the coat) and began in-depth studies with her colleagues Eduardo Eizirik and Jonas Lescroart.

By comparing the DNA of two specimens found in Bolivian forests with that of 38 similar species belonging to the genus Leopardus, researchers concluded that they had encountered a new variety, the first to be discovered since the pampas cat in 1923. In light of these recent findings, the researchers have also stated that the family of South American ‘tiger cats’ comprises a total of five species, as opposed to the two previously recorded.

At present, information on the tilcayo is extremely limited. The feline measures less than 50 cm and weighs 1.5 kg, and has very long whiskers compared to domestic varieties. With regard to its ethology – that is, its behaviour in the wild – the initial hypothesis is that the tiger cat is predominantly nocturnal and feeds on small rodents. Lescroart stated that the tilcayo is classified as a ‘cryptic’ species, meaning it cannot be identified solely by its physical appearance. The word ‘tilcayo’, used by local people to refer to the species, does not appear to have any particular meaning in Spanish or Quechua.

The discovery was published yesterday in the scientific journal Current Biology. In the study, the authors emphasise that the conservation status of the feline is unknown, and therefore preserving the habitat of the Bolivian Yungas will be crucial for the future of the ‘new cat’.

To understand what has been done, we need to start with a real problem in contemporary scientific literature. An article on computational biology may describe a new method with sufficient precision for an expert to understand its principle, whilst still leaving the reader with a considerable amount of work to do before that method can be used: one must find the programme written by the authors, install the libraries on which it depends, understand what data must be provided and in what format, reconstruct the sequence of operations from the tutorials and, finally, adapt everything to one’s own situation. In other words, the paper conveys knowledge, whilst an increasingly significant part of that knowledge also exists in the form of code and procedures that need to be put back into practice.

Paper2Agent aims to automate precisely this step. The system reads the article and its accompanying material, identifies the repository containing the code, sets up an environment in which that code can run, examines the tutorials provided by the authors, and attempts to transform the operations it finds into general functions. These functions are then made available via MCP (Model Context Protocol), a standard that enables a language model to utilise external resources and programmes through a common interface: rather than asking the model to write the code required to perform an analysis each time, it is, so to speak, provided with a toolbox of pre-prepared tools, each with a name, specific inputs and an expected output.

The most interesting aspect of the architecture is that another agent tests the functions obtained using the same examples as those used by the paper’s authors and checks that they produce the expected files, that the numerical results match within a defined tolerance, and that even the figures are sufficiently similar to the reference images. If anything fails, the system attempts to correct it; after six unsuccessful attempts, that function is excluded from the final agent. The result is an MCP server containing only those tools that have passed this verification, along with the text of the article, supplementary materials and instructions describing certain workflows, i.e. sequences of scientific operations.

At this point, a second LLM comes into play, acting as the interface with the researcher. A question can be phrased in natural language, and the model decides which tools to use, with which parameters and in what order; it then carries out the operations via the MCP server and returns the result. In the case of AlphaGenome, a model that predicts the effects of DNA variants on gene regulation, Paper2Agent has produced twenty-two validated tools; the user can therefore ask what a particular mutation might do in a certain cell type without having to learn how to use the original software directly.

This part functions sufficiently to render the work significant, regardless of the surrounding rhetoric. The authors took one hundred computational biology articles from bioRxiv without selecting them based on the quality of the available code. In 74 cases, the process resulted in the construction of a working agent; in the remaining 26, either executable code, necessary data or models were missing, or there were unresolvable software dependencies and programmes that were too closely tied to the specific original example. The 74 usable articles yielded 599 potential tools, 593 of which passed automated verification. This is already sufficient to correct the first exaggeration in the formula chosen by Nature: at least in its executable component, Paper2Agent does not turn ‘every article’ into an agent at all.

Out of 300 questions drawn from the tutorials in the 74 articles, Paper2Agent with Claude Sonnet 4 provided the answer deemed correct in 91.2 per cent of cases, whilst Claude Code, which was provided with only the article and the original programme repository, achieved only 80.3 per cent; even a later version of the model, Sonnet 4.6, reached 86.3 per cent. In ten computational studies from other fields, the percentage reported by the authors reached 98.1 per cent across 42 tasks. For 26 articles that were predominantly experimental or focused on new data—from which it was not possible to derive actual executable tools—Paper2Agent instead constructed a structured collection of text and supplementary materials, achieving 89 per cent on 100 summary questions, compared with 82 per cent for a model permitted to consult the paper directly.

The difference has a fairly simple explanation. Asking a language model every time to read through thousands of lines of code, understand how to install a scientific programme and improvise the correct procedure on the spot leaves enormous scope for errors. Paper2Agent carries out much of that work just once, tests what it has built and then provides the model with a smaller set of operations that already work. Experiments in which the authors remove certain components of the system confirm that it is precisely the verification phase that contributes significantly to the final performance. The technological solution, therefore, exists: it is possible to transform a considerable proportion of the software accompanying scientific literature into functions that can be queried using natural language, thereby increasing the likelihood that they will be executed as the authors intended.

From here, however, the authors and Nature take a much more ambitious leap. The authors speak of articles becoming “competent entities capable of execution and dialogue”; they envisage that, in future, an “agent availability” section will be added to the usual statements on the availability of data and code; and they describe communities of agents capable of autonomously linking methods and results from different disciplines. In one experiment, they have agents derived from AlphaGenome and from two papers containing data on T cells interact, identifying GPR137 as a candidate to explain the effect of a genetic variant associated with psoriasis; the agent proposes ten strategies for moving forward, and a researcher selects the one that is ultimately applied to the data. The authors themselves, in the final version, make it clear that the generation of hypotheses and the interpretation of mechanisms remain under human control.

This clarification came at the end of a particularly instructive peer review. A reviewer observed that the accuracy of the tools is measured by comparing them with the very code from which they were derived: if they faithfully reproduce the published implementation, they receive a high score, even though this does not prove that that implementation represents the only scientifically valid analysis, let alone that it is scientifically correct. The authors accepted the objection and, in the published version, explicitly state that, in open problems, these benchmarks should be interpreted primarily as measures of ‘faithful execution’ – that is, faithful execution of the published method – rather than as measures of general analytical validity.

This distinction is essential. If there is a questionable statistical choice in the original code, reproducing it perfectly does not make it any better; if there are two scientifically valid procedures and the benchmark result uses one of them, an agent that correctly chooses the other may even be deemed ‘wrong’ in the benchmark. Paper2Agent verifies with remarkable effectiveness that a machine is able to replicate what the paper teaches it to do. The validity of what the paper teaches remains a scientific matter.

This also highlights a deeper issue, which concerns the very strongest term used by the authors: reproducibility.

Traditional scientific code may contain errors – and often does – but it possesses one fundamental property: once the programme, data, parameters and execution environment have been fixed, we at least have the possibility of specifying exactly what is to happen. Paper2Agent commendably attempts to ‘freeze’ this part of the process, because it builds tools, tests them and, once they have passed verification, ‘locks’ them in place. However, the moment those tools are entrusted to a linguistic agent, the scientific chain once again incorporates an element of a different nature.

A large language model does not store a sequence of rules of the type ‘if you find A, perform B’. Given what it has received up to a certain point, it calculates a probability distribution over possible continuations, and the next step is generated from that distribution. For this reason, the referees themselves asked the authors to repeat the experiments several times: they explicitly state that LLMs are stochastic. The authors therefore ran the same benchmarks five times and stopped reporting just a single percentage figure, instead introducing means and variability across the replicates. For AlphaGenome, for example, the 100 per cent initially emphasised for tasks derived from the tutorials became 98.7 per cent with a standard error of 1.3 points.

The word ‘probabilistic’ requires clarification. It does not mean that every response must necessarily change, nor that it is impossible to make a model very stable by fixing the version, the generation settings and the infrastructure. Rather, it means that the machine to which we are entrusting part of the procedure is not an explicit sequence of methodological choices written by scientists; it produces its own decisions based on a learnt function, and when the generation process allows for sampling, different trajectories may emerge even when faced with the same task. Furthermore, if the model is provided as a remote service, the issue is compounded by updates to the model and the infrastructure running it.

As long as the LLM merely needs to understand that the user has requested tool number seven and pass a parameter to it, the consequences are limited. Paper2Agent already entrusts it with something more interesting. In open-ended tasks, the model must independently formulate a plan, combine different tools and synthesise the results into a biological conclusion; in the example with Scanpy, the prompt instructs the agent to inspect the data prior to analysis in order to select appropriate parameters. It is precisely these decisions that, in a standard ‘Materials and Methods’ section, we would like to know about in order to understand how a result was produced.

Let us suppose, then, that we have ten perfectly reproducible tools. The final result may depend on the choice to use seven rather than eight of them, on the order in which they are called, on the value assigned to a threshold, or on the interpretation of an intermediate result. If those choices are made by the linguistic model, we have made the individual tools reproducible and have placed a probabilistic machine in the position where decisions are made about which tools to use.

It is possible to record everything. By preserving the conversation, the calls made, the parameters and the outputs, another researcher could retrace that particular analysis exactly. We would have achieved reproducibility of the path already taken – a sort of perfect flight recorder. A different question remains: if the same problem were presented to the same agent again, would it have chosen that same path? If the answer depends on the outcome of a probabilistic generation, the log allows us to replicate the trajectory produced on one occasion, whilst the rule that generated that trajectory continues to differ from the explicit protocol to which science has accustomed us.

The difficulty persists even if the generation process were to be made technically deterministic. A model that always produced the same decision given the same input would resolve the issue of material repetition, whilst leaving open that of scientific justification: why did it choose that parameter? Why did it favour that test? What methodological assumption links the observation to the choice? The answer generated retrospectively by the model does not necessarily coincide with the computational cause that produced the decision. In a scientific article, however, those assumptions form part of the subject matter that must be subjected to scrutiny by other researchers.

Paper2Agent therefore contains two ideas that deserve different fates. The first is already a strong one today: transforming code, data and tutorials scattered around an article into a set of verified, versioned and easily executable tools could genuinely improve the reuse of computational literature, and the fact that 26 of the 100 repositories examined did not reach completion could itself become an excellent practical test of the quality with which we publish scientific software. The authors go so far as to suggest precisely this possibility.

The second idea – that of a paper transformed into a ‘virtual author’ and then into a member of a society of scientist-agents – requires much greater caution. A corresponding author is not the voice interface of their own repository: they are aware of the decisions that precede the code, know which alternatives have been discarded, and should be capable of assuming intellectual responsibility for the interpretation. The Paper2Agent agent knows what is presented to it through the article and its artefacts and uses an LLM to decide how to query that material. The similarity to an author therefore concerns the conversation, and far less so the epistemology.

A scientifically more robust version of the idea is already evident within the paper itself. The LLM can be used to transform a human query into a workflow, find the relevant code, set up the environment and even suggest possible strategies. However, when that process produces a result that is to be included in the scientific literature, the workflow could be ‘frozen’ into an object independent of the LLM: specific functions, code version, uniquely identified data, chosen parameters, sequence of operations and expected results. At that point, another researcher could re-run it without asking a model to decide anew how to proceed.

Artificial intelligence would thus play a role very similar to that of an extremely sophisticated compiler: translating a scientific intention expressed in human language into an explicit procedure which, before becoming part of the published result, can be inspected and fixed. Allowing the LLM to operate within the protocol every time the paper is queried, on the other hand, leaves open the possibility that the method may change alongside its execution.

For almost four centuries, scientific publishing has sought – with rather limited success – to separate the result from the person reporting it, to the extent that others can verify how it was obtained. Paper2Agent presents a plausible way of making articles far more dynamic than the PDFs we still use today, finally enabling text, code and data to work together. However, this promise only becomes scientifically interesting on condition that the ‘active’ article retains a property that the old ‘passive’ article at least sought to guarantee: when faced with a result, we must be able to reconstruct a procedure, and that procedure cannot depend, in its decisive aspect, on whichever response a probabilistic model chose to generate on that particular day.

The stakes are therefore much higher than the discovery of some new microorganism, because the origin of life seeks to explain how, on the early Earth, chemical processes produced the first systems capable of reproducing and evolving; whilst alongside this there is a second question regarding origins, less well known to the general public but almost as important: how that life – which for an immense period of time consisted exclusively of relatively simple cells – managed to produce the complex cell from which almost everything we now call complex life is descended.

For around two billion years after the emergence of the first life forms, only organisms consisting of cells without a nucleus existed on Earth; bacteria and archaea achieved – and still maintain today – an extraordinary variety of forms and metabolic processes, dominating many of the planet’s environments, whilst their cellular organisation remains profoundly different from our own. At a certain point in Earth’s history, however, a larger cell appeared, equipped with a nucleus in which to store DNA, internal membranes capable of creating specialised compartments, and a network of filaments – the cytoskeleton – with which to change shape, organise its interior and move materials from one region to another; mitochondria also emerged—the structures that still today enable our cells to harness the energy contained in nutrients with great efficiency.

That was one of the greatest transitions in the history of life, because multicellularity would emerge much later and along different evolutionary lines, giving rise separately to animals, plants, fungi and other multicellular organisms; yet all these histories depend on the earlier emergence of the eukaryotic cell. Before an animal could be formed, it was in fact necessary to have cells capable of growing large, of finely controlling their own shape and contents, of moving molecules within them and of utilising energy efficiently; eukaryogenesis—that is, the process through which this new type of cell arose—therefore constitutes, in a very concrete sense, the origin of complex life.

One of the latest organisms that could help us reconstruct that transition is called Nerearchaeum marumarumayae and was recovered from the microbial communities associated with the stromatolites of Shark Bay in Western Australia; stromatolites are layered mineral structures built up by the activity of microorganisms and have become famous because similar forms have left some of the oldest traces of life on Earth. After years of work, the researchers have succeeded in obtaining a culture in which Nerearchaeum accounts for around 89 per cent of the cells, whilst the rest of the culture is dominated by a bacterium, Stromatodesulfovibrio nilemahensis, and it is precisely the relationship between the two that appears to be an essential part of the story.

Nerearchaeum produces certain small molecules that the bacterium can utilise, whilst the bacterium is capable of producing compounds on which the archaeon appears to depend; thus, the two organisms do not simply share the same environment, but intertwine part of their survival through each other’s metabolism. When the researchers observed them using cryo-electron tomography – a technique that allows frozen cells to be reconstructed in three dimensions almost instantaneously – they saw that the archaeon and bacterium may be connected by extremely thin tubular structures. This observation is significant because it reveals something that has long occupied a central position in reconstructing the origin of eukaryotes: the possibility that a very close metabolic relationship between an archaeon and a bacterium preceded their integration into a single cell.

This does not mean that Nerearchaeum and its partner are re-enacting, under our microscopes, what happened two billion years ago, because modern organisms have undergone two billion years of independent evolution and are not living fossils; however, they do demonstrate that such relationships are biologically possible and finally allow us to study their actual form, rather than merely inferring it by comparing the genomes of modern organisms.

The reason we are specifically looking for a pair consisting of an archaeon and a bacterium stems from one of the most spectacular discoveries in 20th-century biology: the mitochondria present in our cells are descended from bacteria which, in the very distant past, began to live permanently inside another cell and which, over the course of evolution, lost a huge part of their autonomy, transferred many of their genes into the host’s DNA and eventually became indispensable components of the new cell. This process is called endosymbiosis, because one organism comes to live inside another and, generation after generation, the relationship becomes so close that it gives rise to a new biological unit.

We still carry traces of this within every one of our cells, given that mitochondria retain a small amount of their own DNA, divide in a manner reminiscent of bacteria, and numerous components clearly reveal their bacterial origin; the event that produced them therefore did not represent the emergence out of nowhere of a new cellular structure, but the result of the evolutionary union between two organisms that, until that point, had followed separate evolutionary paths.

We know quite a lot about one of the two protagonists, because the future mitochondrion was derived from a bacterium belonging to a large lineage called Alphaproteobacteria, whilst the identity of the other protagonist has been much more difficult to piece together. We knew it was linked to archaea, because many of the fundamental systems by which eukaryotes copy DNA, produce RNA and build proteins have close relatives within that very group, but we were missing something that resembled our ancestor closely enough to allow us to imagine what sort of cell the future mitochondrion had encountered.

In 2015, the picture changed when, by analysing DNA recovered from seabed sediments, researchers reconstructed the genome of a previously unknown organism, named Lokiarchaeum; in the years that followed, other related groups emerged, to which names taken from Norse mythology were assigned, including Thor, Odin and Heimdall, and the entire group came to be known as Asgard.

Their position in the evolutionary tree immediately attracted attention, as eukaryotes were found to be closely related to these archaea, and a second surprise followed shortly afterwards, one that was even more significant from a biological point of view: their genomes contained genes for proteins similar to those that eukaryotes use to organise the cell, alter its shape and control its membranes, to such an extent that some of those proteins had, until then, been regarded almost as a hallmark of the eukaryotic cell.

The finding altered the problem, because it became possible to imagine that the cell from which we are descended was already more sophisticated than previously thought before encountering the bacterium destined to become the mitochondrion; yet an enormous limitation remained, given that for years the Asgard were known almost exclusively through their DNA, and attempting to reconstruct their behaviour was, to some extent, like trying to understand an organism by referring to its genetic manual without ever having seen it.

The next breakthrough came when certain research groups finally succeeded in cultivating them, and the first success was Prometheoarchaeum syntrophicum, achieved after more than ten years of attempts using marine sediments. It is a tiny cell, grows extremely slowly and is metabolically dependent on other microorganisms, whilst thin, branched membrane extensions protrude from its surface.

These structures suggested a fascinating hypothesis, according to which an ancient archaeon might first have established a close metabolic relationship with a bacterium and then gradually encircled its partner with its own cellular extensions, eventually engulfing it; the bacterium inside would have continued to provide energy and other metabolic benefits, whilst losing its independence until it became a mitochondrion.

It is merely one of several possible models and many details remain open to debate, but the significance of Prometheoarchaeum lay elsewhere, because it demonstrated that an archaeon belonging precisely to the group closest to eukaryotes could indeed live in very close dependence on other microorganisms and produce large expansions of its own cell surface, thereby taking two characteristics required by certain scenarios of eukaryogenesis out of the realm of pure theoretical possibility.

Then came Lokiarchaeum ossiferum, and upon observing it in three dimensions, the researchers discovered that a network of filaments runs through the cell and its protrusions, formed by a protein related to actin – one of the main proteins in our cytoskeleton, which is essential for enabling eukaryotic cells to change shape, move and organise their internal structure. The discovery thus showed that at least some elements of the cellular machinery we typically consider to be eukaryotic already existed, in a simpler form, in our archaeal relatives.

Nerearchaeum now adds a further element, as it allows us to directly observe an Asgard linked to a bacterium through a metabolic relationship and close physical contact; taken together, these organisms are transforming a problem that, until a few years ago, was addressed almost exclusively through the comparison of genetic sequences. Researchers can now begin to see the cells, observe how they build their membranes, understand which partners they need, and verify what those proteins – which share an evolutionary history with our own – actually do.

Reconstructions of ancient genomes are also altering our understanding of the event, because all eukaryotes existing today are descended from a common ancestor that biologists call LECA, an acronym for Last Eukaryotic Common Ancestor, and LECA was not a rudimentary cell at the start of the transformation, but already possessed mitochondria and most of the fundamental components of a modern eukaryotic cell. When we encounter LECA in the evolutionary reconstruction, therefore, much of the revolution has already taken place and the real enigma lies in the preceding period, between the archaeal ancestor and LECA.

A recent study has sought to establish which of the two ancient partners provided the various cellular systems already present in the common ancestor of eukaryotes, and the picture that emerges is one in which many of the structures necessary for the organisation of the cell appear to have their roots in the Asgard lineage, whilst the contribution of the bacterial ancestor of the mitochondrion is particularly significant in systems related to energy production and management.

If this reconstruction continues to hold true, the origin of the eukaryotic cell will appear less and less as the encounter between a very simple archaeal cell and a bacterium capable of producing energy, and increasingly as the integration of two evolutionary histories that had already developed different capabilities: on the one hand, an archaeal cell that was acquiring the means to control its shape and membranes; on the other, a bacterium equipped with an energy metabolism destined to become the biochemical heart of the new cell.

It remains to be determined when and in what order the nucleus, the internal membrane systems, the more elaborate cytoskeleton and many of the other eukaryotic characteristics appeared; it also remains to be understood to what extent the emergence of the future mitochondrion enabled subsequent innovations, providing the cell with a previously unattainable level of energy and the means to manage it. There are still differing models on these points, and it is the Asgards who are, for the first time, providing experimental data with which to test them.

The significance of this issue lies in the scale of the event, because the origin of life seeks to explain how something capable of evolving first appeared on Earth, whilst the origin of the eukaryotic cell seeks to explain how, much later, that life acquired a new organisational structure that would make possible almost all the biological complexity visible around us.

For billions of years, the Earth was a planet of microbes and then, on a single occasion as far as we can reconstruct, two profoundly different cellular lineages eventually merged to produce a completely new type of organism; all the cells in our bodies, those of a sequoia, those of a fungus and those of a whale are descended from that innovation. Understanding how this happened therefore means reconstructing one of the very few stages in natural history without which the living world as we know it would never have existed.

And this is what makes the tiny, extremely slow archaea discussed in today’s issue of Nature so important: they do not carry the answer already encoded within them, nor do they represent ancestors frozen in time; rather, they allow us, for the first time, to observe at close quarters cells belonging to the branch of life closest to our own as they build complex structures, depend on other microorganisms and establish such close relationships with bacteria that we can glimpse, in their biological potential, some of the steps through which simple life was able to evolve into complex life.

Whenever an indicator reaches a high value, the incentive to obtain it by separating it from the reality it is supposed to represent also increases. In the case of publications, this practice has now taken on an industrial scale. There are organisations that produce fake or recycled scientific articles, find them a publisher, and sell a place on the list of authors for a few hundred dollars. They are known as ‘paper mills’, and their end product is not simply poor-quality scientific work: it is a publication designed to allow someone to include research on their CV that they have never actually carried out.

On 6 September, the science desk at ERR, Estonia’s public broadcaster, set out to see just how difficult it would be to break into this market. The answer came within half an hour. Journalists at Novaator, in an investigation written by Johannes Peetsalu and edited by Jaan-Juhan Oidermaa, created a fictional IT specialist named John Robert Steezy, gathered the contact details of seven particularly active sellers on Facebook, and messaged them on WhatsApp explicitly asking whether it was possible to buy an article byline. Five replied within thirty minutes. During the same period, more than twenty offers arrived on the fictitious researcher’s phone, with prices ranging from two hundred to eight hundred dollars.

There was no need to frequent underground forums, know intermediaries or learn any jargon reserved for insiders. Sellers openly advertised their services in Facebook groups with names such as ‘Free Scopus publication and research help’, using posters generated by artificial intelligence and inviting potential customers to continue the conversation on WhatsApp. Behind the façade of editorial consultancy lay a standard commercial catalogue from which one could choose the journal, one’s position amongst the authors, turnaround times and price.

The most expensive offer received by ERR was for eight hundred dollars and concerned an article intended, according to the seller, for the Journal of Intelligent Decision Making and Information Science. Publication was promised within twenty days. This timeframe is incompatible with what is normally required for a rigorous editorial assessment, the search for reviewers, peer review, any revisions to the manuscript and subsequent editorial processing. At the other end of the price range, for two hundred dollars, the fake computer scientist was offered the opportunity to be listed as an author on an article in fisheries science – a field, of course, completely outside the area in which the supposed researcher was an expert.

Even the customer service department had reached a remarkable level of professionalism. One of the sales representatives spoke to the journalist on the telephone, explained the publication procedure and pricing system to him in English, and made it clear that he would only have to pay once the manuscript had been accepted by the journal. The scam was therefore presented with a guarantee that shifted a substantial portion of the commercial risk onto the sales representative. Within a few hours, all the conversations had reached the same stage: all it would have taken was an email for one of the manuscripts to be sent to a journal listing John Robert Steezy as one of the authors – a man who does not exist and who, naturally, had not produced a single piece of data contained in those papers. At that point, the editorial team halted the experiment. The salespeople continued to write for days.

The submitted manuscripts were plausible enough to pass at least a cursory reading. On closer inspection, however, they revealed nonsensical figures and bibliographical references inserted haphazardly – signs of a process primarily concerned with reproducing the formal appearance of a scientific article. This is precisely the commercial requirement of a ‘paper mill’: the text must not contain any discovery; it must merely appear to be a publishable piece for long enough to pass the available checks.

To achieve this, factories now have access to various sources of raw materials. Some of these can be generated using artificial intelligence systems. Another part comes directly from existing scientific literature. According to ERR’s findings, little-known works or those published in languages that reduce the likelihood of recognition are also selected; these are then automatically translated and modified in terms of wording, tables and figures to a sufficient extent to make it more difficult to identify plagiarism.

Lauri Laanisto, a tenure-track professor of macroecology at the Estonian University of Life Sciences, had the opportunity to observe the process at first hand. In the spring of 2026, he discovered that a research paper in which he had participated had been submitted to a journal, slightly altered, under another person’s name. The discovery was a stroke of luck: a few months earlier, Laanisto had presented those very same results, including the figures, at a conference in Poland attended by the editor-in-chief of the journal that would later receive the plagiarised manuscript. When the editor came across the new article, he recognised material he had already seen.

The process of altering the text was straightforward and, for that very reason, quite instructive. The title and wording of the text had been changed; an image that was originally in colour had been converted to black and white; the results and scientific structure remained those of the original study. The manuscript listed as its author a person affiliated with a Chinese university, of whom Laanisto was unable to find any trace either on the university’s website or in academic databases. Nor is it certain that whoever produced the forgery needed to proceed as far as publication. Laanisto observes that, in some evaluation systems, a PhD student or researcher facing a deadline may only need to demonstrate that they have submitted a manuscript to a journal. In that case, the submission receipt already constitutes the ‘product’ purchased.

There remains, then, the issue of the checks that are supposed to prevent such material from becoming scientific literature. Here too, the industry has developed appropriate solutions. One of the techniques described by ERR involves proposing non-existent reviewers to the journal, linking them to email addresses controlled by the very organisation that produced the manuscript; when the publisher sends the request for peer review, the text thus returns to the hands of those responsible for getting it published, who can then fabricate a favourable assessment under a false identity. There is also documented evidence of attempts to bribe editors and of operations capable of taking control of parts of the editorial process.

The scale of the phenomenon makes it difficult to dismiss it all as merely a colourful fringe of poor-quality scientific publishing. An analysis published in PNAS identified 29,956 articles suspected of originating from ‘paper mills’ and also present in OpenAlex. At the time of the analysis, 8,589 of these – 28.7 per cent – had been retracted. Based on the observed trend, the authors estimate that only about a quarter of these papers will ever be retracted and that only an even smaller fraction will end up in journals subsequently excluded from the main indexing systems. A large amount of presumably fraudulent material is therefore set to remain part of the accessible literature, where it can be found, cited and incorporated into subsequent works.

Here, the harm takes on a different nature. Anyone who buys authorship of an article gains an illegitimate advantage on their CV and competes with researchers who have genuinely produced their own work; once published, the purchased article simultaneously enters the knowledge base that other scientists use to formulate hypotheses, design experiments and synthesise available knowledge. If it relates to biomedical disciplines, it may end up in literature reviews and, through subsequent stages, even contribute to the development of clinical guidelines. A person’s falsified CV thus becomes a problem for the reliability of collective knowledge.

Laanisto points out that the buying and selling of bylines has been going on for at least twenty years. In the past, this could even involve genuine articles that had already passed an initial review and in which, prior to final publication, a place amongst the authors was sold. The new development documented by the Estonian experiment concerns the scale and ease of access: the customer no longer needs to be in the right network at the right time, as they can contact a seller directly on social media and, within minutes, receive a selection of articles that are already available.

The demand fuelling this trade also stems from the way in which many institutions have chosen to evaluate research. Laanisto uses a very simple comparison: for a politician, it is votes that count; for a researcher, it is publications. The comparison becomes particularly apt when the number of papers is a direct factor in the conditions for securing a post, retaining it or receiving funding. In some countries, there are even cash rewards linked to publications; according to Laanisto, it is chiefly in contexts of this kind that ‘paper mills’ have so far found the most favourable ground, whilst he is not aware of any cases of Estonian researchers having made use of them.

However, it would be too simplistic to attribute the entire phenomenon to the pressure of ‘publish or perish’. That pressure creates demand, whilst for an industry to emerge, there must be a reasonable likelihood that the purchase will be successful. Paper mills can thrive because a publication retains high value as a certificate of productivity even when the mechanisms that should guarantee who produced it and how it was produced become compromised. Quantitative assessment, insufficient editorial checks and the lack of consequences following discovery thus end up assigning a market price to something that the scientific system continues to treat as proof of merit.

At present, anyone found to have purchased an article, duplicated data or published material stolen from other researchers may face local consequences without necessarily incurring a sanction that would follow them throughout the academic system. A formal exclusion from access to public funding, however, would alter the risk assessment. At present, that calculation can be remarkably favourable: a few hundred dollars for an extra publication, an uncertain likelihood of being discovered, and consequences that may remain limited even when the fraud comes to light.

ERR’s experiment, very similar to one I carried out myself about 10 years ago, highlights above all just how commonplace this market has become. Seven telephone numbers found on Facebook, five replies in half an hour, more than twenty articles to choose from, two hundred dollars to take on a role as an IT specialist in a fisheries project, eight hundred for a promise of publication within twenty days. Behind these figures lies a business that now understands academic demand, tailors its services to the metrics used to evaluate researchers, and supports the client right through to the acceptance of the manuscript. When buying documentary evidence of having conducted scientific research becomes easier than actually carrying out the research required to produce it, the problem no longer concerns only those who cheat, but the value we are still willing to attribute to that evidence. When will we put an end to bibliometrics?

Fictitious experimental data have always accompanied research, and until recently were recognised as such because the procedure used to fabricate them left identifiable traces. The most widely studied type concerns noise, that is to say, the unintentional variability inherent in every real measurement, arising from the instrument, the sample and the hand of the person working at the bench. Anyone who sits down at a desk to fill in a table of measurements that were never actually taken generates noise that is far too ‘polite’, because our intuitive idea of variability coincides with that of an orderly dispersion around the expected value: values that repeat at short intervals are avoided, as they seem suspicious; the final decimal places are spread out evenly; and the results are kept reasonably close to the mean. Authentic noise is far less disciplined in nature and accumulates improbable coincidences, values that are extremely far removed from the others, and clusters of almost identical results that no forger would dare to transcribe.

Noise is merely one of the traces that the method of production leaves on the final result. Those who duplicate a set of measurements and resize it to suit a second experiment leave sequences that reappear several lines later, whilst those who rely on a spreadsheet to generate data produce distributions that clearly reveal the type of generator used. The reported means may prove arithmetically incompatible with the number of subjects and the scale used to measure them, and the agreement between the experimental data and the theoretical model may appear so perfect as to become implausible, as is the case with the initial characteristics of the arms in a randomised trial when they resemble one another more closely than a random draw would suggest. In short, every shortcut taken during the fabrication process leaves a statistical artefact somewhere within the figures, and it is upon this phenomenology that the edifice of post-publication scrutiny has been built over the last thirty years – a process that hunts down the fingerprint left by the forger’s methods far more than the content of the forgery itself.

A generative model does not take any of those shortcuts, because it samples directly from the pattern that real data has shown it millions of times during training, and produces tables featuring credible noise, consistent internal correlations, and missing values distributed as they would be in the real-life setting of a hospital ward, alongside sequencing reads, mass spectra, electronic health records populated with patients who were never born. The forgery emerges devoid of the tell-tale signs we were looking for and equipped with just the right imperfections, and the very foundations upon which our verification tools rested crumble just as the cost of production plummets to a few dozen euros per item. The falsification of scientific images using generative AI—which I documented last year with Angelo Parini in the American Journal of Hematology—offers only the most striking example of a trend affecting the entire chain of experimental evidence, from the raw data output by the instrument to the summary table that reaches the reviewer’s eyes.

The race between those who fabricate data and those who expose it is, moreover, a foregone conclusion, for the structural reason that every whistleblower becomes training material for the next generation of fraudsters. Asking the scientific community to chase after fraudulent data means condemning it to lose ground with every model update, with the depressing consequence that the only guaranteed result will be a constant stream of scandals to report.

The approach now being adopted in the clinical setting turns the issue on its head and certifies the original data at the exact moment it is produced. In regulated research, this practice is already standard: the system records data via an unalterable audit trail; the raw data is generated already signed and time-stamped; the inspector traces it back to the source and verifies the chain of custody leading from the patient’s record to the published table. Bringing the same architecture into the preclinical laboratory is a matter of straightforward engineering, given that the equipment used there can sign the file upon acquisition just as naturally as it currently records the date and parameters on it, and a journal can require, alongside the manuscript, the cryptographic fingerprint of the raw data from which each table is derived. At that point, fabrication would require the falsification of the chain of custody, which is an act of a completely different nature, one that is traceable and punishable under criminal law.

All this, however, remains a band-aid applied to an organism that continues to produce the disease, because the root of the problem lies elsewhere and we have known about it for decades. The demand for fake articles is generated by an evaluation system that measures people by counting their output, and which has turned the number of publications and the journal hosting them into the currency with which competitions, promotions and funding are bought. The San Francisco Declaration on Research Evaluation has been denouncing this drift since 2012; universities sign it enthusiastically yet continue to compile their internal rankings in exactly the same way, whilst entire national systems have, by law, tied academic careers to numerical thresholds that no committee can ignore. As long as researchers face the choice between publishing ten mediocre papers and remaining stagnant with two serious ones, the ‘paper mill’ industry will have a stable and growing clientele, and any technical barrier will merely serve to drive up the price of the forgery.

These are two measures that only make sense when taken together, because data certification makes fraud difficult, whilst the reform of the evaluation system renders it pointless. Neither is particularly suited to a front-page headline, and this is perhaps why the public debate touches on them only in passing, before returning to the story of the individual case – the professor who has been exposed, the journal forced to retract the article.

Investigative journalism serves a useful purpose when it brings the issue back to those who have the means to tackle it. In this area, the news has been circulating for some time and, with every new report, produces the same wave of outrage, which invariably recedes, leaving the shore exactly as it was; and the risk, when we stop at mere indignation, is that we hand those who are already sceptical of medicine yet another argument, whilst leaving the machine that churns out false data intact, along with the incentives that fuel it.

Changing those incentives, and making those who cheat pay a price proportionate to the damage they have caused, are tasks that fall to society as a whole, because no laboratory can rewrite the rules of competitions and no journal has the power to impose a penalty. Each of these decisions, however, rests on technical evidence capable of standing up to an appeal or a court of law, and such evidence can only be established by genuine science, through forensic analysis of the figures and the chain of custody linking the raw data to the published table. The conclusion may appear paradoxical only to those observing the matter from afar: to rid ourselves of corrupt science, we need more science – in far greater quantity and of far higher quality than we currently devote to the problem.

A few weeks earlier, Andreas Thom, from Dresden University of Technology, had asked a similar question after the Astra model had constructed the first non-sofic group with a decisive step based on a 2019 article he had written with Gábor Kun: Thom had been discussing precisely that issue with ChatGPT for months, and the response he received from an OpenAI researcher – ‘that did not happen’ – as he reports it, ruled out direct access to the conversations and remained silent on the training. Neither Buckmaster nor Thom claim that the transfer took place, yet both have put their finger on an issue that goes far beyond their individual cases.

In August, a few days after Astra’s announcement, Francesco Fournier-Facio, in Cambridge, realised that OpenAI’s press release had omitted the articles on which that result was based, to the extent that the company had amended the text: the sources had been published, and a specialist was able to identify them within a few days. When the potential source is a private conversation, however, the same process becomes impractical for anyone outside the company and can prove extremely difficult even for the company itself, because in a language model, training modifies billions of parameters and what the system has learnt typically remains a capability without a verifiable origin; indeed, determining which fragment of data influenced a particular response remains an open research problem. OpenAI can verify whether certain conversations have entered its dataset, whilst it would be far more difficult to reconstruct the path that would have transformed them into an insight generated by the machine.

Psychologists have a name for this sort of bewilderment. In 1902, Carl Gustav Jung noted that a passage from Thus Spoke Zarathustra mirrored almost word for word a page by Justinus Kerner that Nietzsche had read as a boy, and he called the phenomenon ‘cryptomnesia’, which cognitive psychology now describes as a source monitoring error, in which the content survives whilst the label indicating its origin is detached. A model trained on its users’ conversations could produce a functionally equivalent phenomenon on an industrial scale, in which no one would open a confidential file, and yet an idea confided to a chatbot might resurface some time later as a suggestion from the machine, without the researcher receiving it, the one publishing it, or the company advertising it knowing where it came from. The de-identification promised by the terms of use leaves the problem intact, because it removes the name whilst retaining the idea – precisely what is of value to a scientist.

The loss of traceability carries such weight because the entire architecture of modern science rests on the principle of priority. Robert K. Merton explained this in 1957: results become part of the public domain the moment they are published, and the scientist’s sole remaining claim to ownership is recognition – the name associated with a theorem or a discovery – so that the rule rewarding the first to arrive only works as long as it is possible to establish who arrived first. This is why, over the centuries, science has devised methods of dating, from the Latin anagram with which Galileo secured priority over the phases of Venus in 1610 to the sealed envelopes that the Académie des sciences in Paris has been receiving since the late 17th century. The chatbot to which a researcher confides an unfinished idea is like a sealed envelope delivered to a recipient who can open it, who issues no receipt, and who, in the meantime, is competing on the very same problems – in a sector where the announcement ‘the machine has solved it’ has become the currency by which a company’s value is measured; and the risk affects the entire sector, albeit with rules that vary from one company to another.

The consequences for the integrity of research stem from this. The European Code of Conduct drawn up by ALLEA, which the Commission recognises as a benchmark for EU-funded projects, defines plagiarism as the use of another person’s work or ideas without due acknowledgement, and the European guidelines on generative artificial intelligence reiterate that researchers remain fully responsible for what they publish. Everything therefore hinges on the author, that is, the link in the chain who has no way of seeing the source: the honest researcher who accepts a suggestion from the model has a duty to acknowledge any hidden contribution by others, yet has no means of knowing whether such a contribution exists; consequently, they might publish, in perfectly good faith, an idea that originated in a colleague’s chat, whilst the entity controlling the channel may be a private company over which these self-regulatory codes do not necessarily exert any binding force.

This leads to the most reasonable request, which is in fact the one put forward by Thom: the onus to shed light on the matter lies with those who hold the data. Companies that claim to have made discoveries can be asked to carry out an audit of the data’s origin, a process that is simple in principle, albeit technically challenging, to verify – using available methods – whether the key ideas appear in user data from previous months and to publicly account for how the verification was carried out; whilst researchers would be well advised to disable training on their own data and perhaps return to a modern version of the sealed envelope, depositing in an archive with a certified date the material they are about to discuss with a machine. In 1976, a New York court ruled that George Harrison had reproduced the melody of ‘He’s So Fine’ by The Chiffons in ‘My Sweet Lord’ and, whilst acknowledging that he had done so unwittingly, held him liable nonetheless, establishing a principle that science has always practised with regard to citations: the copier’s lack of awareness may exonerate their intentions but leaves the debt to the author intact. This should apply all the more so to models that retain what they have learnt even when the user subsequently cuts off the flow of their own data, and which, just like Nietzsche, retain no memory of the source.

From that point onwards, the interview becomes a lengthy exposition of extraordinary medical claims, almost always presented without any verification of their scientific basis. Roia recounts that his Russian mentor, Eugeni Velkhover, was able to detect breast cancer from the iris six years before diagnosis and predict a heart attack two years in advance. He adds that he has learnt to recognise hereditary conditions dating back as far as six generations in the iris, based on statistics allegedly compiled from data on one million people. The Corriere reader receives this information and carries on reading. No one explains to them that iridology has long been the subject of controlled studies and that a systematic review concluded as early as 1999 that its diagnostic validity was not supported by the available evidence. A review published in June 2026 reaches the same conclusion after re-examining biological plausibility, diagnostic accuracy and reproducibility: iridology cannot be upheld as a diagnostic or screening tool.

The issue becomes particularly serious when the interviewee attributes specific therapeutic or diagnostic capabilities to these practices. Voll’s electroacupuncture is presented as a technique that can even be used during pregnancy. There are controlled trials on electrodermal devices derived from that tradition. A randomised, double-blind study published in the British Medical Journal subjected thirty people to more than 1,500 measurements and tested whether the test was capable of distinguishing between people with allergies and those without. It failed. The authors concluded that the method could not be used to diagnose those allergies. This kind of information is completely absent from the Corriere interview.

Then comes hydrocolon therapy. Roia describes individuals said to have retained faecal matter in their bowels for six years, sometimes for twelve, eighteen or twenty-four, lodged in the ‘recesses of the colon’, and claims that this matter becomes visible as it passes through the device’s transparent tube. It is a modern version of the old theory of intestinal auto-intoxication, upon which a branch of Western alternative medicine has been built for decades. Reviews of colon irrigation have found a lack of solid evidence to support the numerous therapeutic claims made by its proponents. One clinical journal went so far as to conclude that these claims were misleading patients. The Corriere, however, publishes the story of twenty-four-year-old faeces without providing the reader with any information on the current state of scientific knowledge.

The same procedure applies to fasting. Roia recounts a ‘record’ of forty-nine days and explains that during fasting, migraines, nausea and tachycardia are allegedly caused by ‘toxins entering the bloodstream’, against which acupuncture is said to be effective. Fasting is a genuine field of biomedical research and, for this very reason, requires precision. The available studies on prolonged fasting describe significant metabolic changes, loss of lean body mass and adverse events; the most recent reviews also highlight the possibility of electrolyte imbalances, orthostatic intolerance and problems with refeeding. Clinical research into fasting can provide specific guidance for specific conditions. It does not provide a blanket licence to turn forty-nine days without food into a therapy based on anecdotal evidence.

The interview eventually turns to anxiety and depression. Roia claims that these can also be treated with his ‘biological medicine’ and justifies this by pointing out that around 90 per cent of serotonin comes from the gut. The quantitative premise contains a fact that is physiologically accurate. The clinical conclusion, however, does not follow from it. Peripheral and cerebral serotonin constitute distinct pools; serotonin produced in the gut does not freely cross the blood-brain barrier. The relationship between the microbiota, the gut and the brain constitutes an important area of contemporary research, with mechanisms far more complex than the oversimplified explanation proposed in the interview. Here too, the reader is presented with a medical explanation of anxiety and depression that lacks the necessary verification.

Lorenzetto expresses some doubts. He describes the term ‘biological medicine’ as ‘vague’, highlights the potential risks of acupuncture, and raises the issues of penicillin, vaccines, antiretrovirals and chemotherapy. These comments remain within the conversation and are addressed in the interviewee’s responses. The journalistic scrutiny required for verifiable health claims would be of a different nature: taking the claim, checking the literature, and reporting the findings to the reader. A newspaper can interview anyone. When it publishes claims capable of influencing a sick person, verifying them is part of the journalistic process.

The most significant question comes near the end: “All right, medicine isn’t an exact science, but who are we supposed to believe?”. A statement such as this incorrectly frames the workings of medicine as a competition between beliefs. Scientific medicine does not ask us to place our trust in one doctor over another. It formulates hypotheses, measures outcomes, compares treatments, records adverse events and revises its conclusions when the evidence changes. The iridologist who claims to see a cancer in the eye six years before diagnosis puts forward a testable hypothesis. Anyone who claims to treat anxiety and depression through a generic ‘biological medicine’ puts forward a testable hypothesis. A major newspaper has the resources to find out what happened when hypotheses of this kind were tested.

Il Corriere also has a Health section, in which it regularly publishes high-quality medical content. It is precisely this fact that makes it difficult to dismiss the case as an inevitable inaccuracy on the part of a general-interest newspaper. The publication itself possesses the necessary expertise to apply basic verification criteria to health-related claims. Evidently, those criteria sometimes cease to apply when medical topics feature in the news, personal interviews or columns.

The Roia incident is, in fact, part of an editorial issue already evident elsewhere. Il Corriere regularly features Franco Berrino’s "Piccole Dosi" column, which appears amongst its opinion pieces. In December 2023, it published an article entitled "Kuzu: how to treat inflammation of the oesophagus, pharynx and intestines", in which acid reflux is described using the yin and yang categories of macrobiotics and it is claimed, on the basis of "clinical experience", that a few weeks’ treatment with kuzu is sufficient to resolve the problem. Another article on constipation states that the best time to have a bowel movement is between six and eight o’clock, when, according to traditional Chinese medicine, ‘the energy of the large intestine’ is at its peak. These articles are published and indexed under the Corriere’s brand, alongside its scientific journalism.

The effect of this editorial system is easy to describe. A claim that lacks sufficient scientific validation nonetheless acquires extremely powerful social credibility. It appears in the Italian daily newspaper which, by virtue of its history, circulation and authority, occupies a central position within the media landscape. It is accompanied by a photograph of the interviewee, details of their degree, the roles they claim to hold and the success stories they recount. The newspaper adds the title ‘one of the leading experts’ in the summary. For a reader lacking the necessary tools to consult PubMed, check diagnostic studies and evaluate a systematic review, this editorial packaging carries enormous weight.

The situation appears even more serious when looking at the Cronache section of Il Corriere dated 8 September. The interview with the doctor discussing iridology, faecal residues dating back twenty-four years and biological treatments for anxiety and depression appears immediately alongside the story of Francesco Gianello, who died aged fourteen from osteosarcoma after his parents had sought treatment from a doctor who followed Hamer’s theories. The experts appointed by the Vicenza Assize Court have just submitted a report stating that the boy could have been saved by following standard chemotherapy and surgical protocols; the trial is ongoing and criminal liability naturally rests with the judges. Il Corriere provides a detailed account of that tragedy and, on the very same page, offers extensive coverage—lacking adequate scientific scrutiny—of another set of medical theories that have no place in evidence-based medicine. The two cases are different and no clinical equivalence can be established. The editorial contradiction remains intact.

For years we have been discussing the damage caused by pseudoscience when it enters institutions. A major daily newspaper is one such institution, because it helps to determine, on a daily basis, which people deserve to be heard and which claims can enter the public discourse with a veneer of respectability. Freedom of speech naturally allows a doctor to advocate iridology, homotoxicology or the existence of faecal matter that has remained in the colon for twenty-four years. Editorial freedom allows a newspaper to interview him. The quality of information is measured by the work carried out between these two freedoms and the reader.

In this case, Il Corriere did not do enough of that work. It turned extraordinary medical claims into interview material, leaving their assessment to the journalist’s next remark. It described a doctor as ‘one of the leading experts’ in relation to ‘biological medicine’ – biological medicine, because by now ‘bio’ is considered good by default – which the interviewee himself combines with practices of deeply varying scientific merit, some of which have failed direct experimental verification. It allowed individual accounts of healing to take up the space where clinical data should have appeared.

When a newspaper of this stature repeatedly makes such a choice, pseudoscience triumphs unchallenged, and the reason is clear: the Corriere della Sera has opened the door to it. Stefano Lorenzetto, the interviewer who wrote a “Dizionario delle citazioni sbagliate”, should know full well what I am talking about.

One can, of course, discuss the German causes of this outcome, just as one might discuss the American, French, Italian or British causes of similar phenomena. There is, however, something more general at play, which unites different societies and concerns the way in which our species makes decisions when it perceives itself to be in danger. Just as the problems requiring in-depth analysis, the ability to distinguish between competing causes, quantitative risk assessment and a willingness to revise one’s conclusions are on the rise, so too is the political demand for rapid answers, simple causes, clear-cut allegiances and leaders capable of making decisions without too many obstacles. When we need rationality most, we become more vulnerable to the shortcuts that undermine it.

An animal must constantly decide whether what it sees, hears or smells poses a danger. The information is almost always incomplete, whilst the potential costs of errors vary greatly. Mistaking a bush swaying in the wind for a predator results in a pointless flight; mistaking a predator for a bush can cost a life – it is no coincidence that predators seek to induce precisely this type of error. When the second error is vastly more costly than the first, natural selection does not necessarily favour the organism that makes the fewest errors overall. Instead, it favours a response threshold shifted in the direction that minimises the most costly error.

Randolph Nesse called this logic the ‘smoke detector principle’. A properly calibrated smoke detector goes off even when we’re simply frying a steak, because the cost of a few false alarms is negligible compared to the cost of failing to detect a fire. Pain, coughing, nausea and many other biological defences operate according to the same asymmetry. Signal detection theory allows us to formalise the problem: as the cost of failing to react to a real danger increases, the amount of evidence required to trigger a response decreases.

An organism that assesses the probability of a catastrophic loss as high therefore has good evolutionary reasons to become less demanding before reacting. False positives increase, because waiting for better information may be more costly than acting immediately on the basis of mediocre information. The resulting response may appear irrational if judged solely on the basis of the accuracy of the prediction; however, in an environment where a false negative entailed a much greater loss of fitness than a false positive, such a threshold might instead have been favoured by natural selection.

In humans, this solution goes hand in hand with an extraordinary dependence on information produced by others. Learning individually is costly: it requires exploration, making mistakes, and investing time; and in certain circumstances, the very mistake needed to learn may be the one that proves fatal. A highly social species can use the behaviour of others as information. If many members of the group avoid a certain area, eat a particular food or follow a specific path, copying them allows one to benefit from knowledge acquired by others without bearing the full cost. Experiments by Thomas Morgan, Luke Rendell, Micael Ehn, William Hoppitt and Kevin Laland have shown that humans do indeed modulate their use of social learning according to, amongst other things, the difficulty of the task, their own confidence and the consensus observed among others, in line with predictions derived from evolutionary models of learning. 

This very solution contains a vulnerability. Copying others works as long as the behaviour of others contains information about reality. If everyone starts copying those around them, an initially poorly-founded decision can spread until it becomes the majority view; at that point, the majority itself becomes a reason to adopt it. The rule that allows a population to accumulate culture can thus produce cascades in which a great many individuals share a choice that very few have independently verified.

The threat also affects the effectiveness of coordination. When the danger is collective, the existence of many incompatible individual strategies can be disastrous, even if some of them are excellent. In a situation where one must flee, defend a territory or resist an attack, it becomes advantageous to quickly agree on a common course of action and to be able to predict that others will do the same.

Patrick Roos, Michele Gelfand, Dana Nau and Janetta Lun have demonstrated, using evolutionary game theory models, that higher levels of threat can foster stronger norms, greater compliance and harsher punishment for deviance, both in problems of cooperation and in those of simple coordination. A thirty-day experiment, published in 2021 by Aron Szekely and colleagues, provided complementary evidence of causality: a higher risk of collective loss led to stronger cooperative norms, which were more resistant to erosion when risk conditions changed.

The exploratory individual, willing to try out new solutions and to diverge from others, can be invaluable when the environment allows for the cost of mistakes to be borne and for the benefits of innovation to be reaped. Under an immediate threat, however, the value of coordination increases, whilst behavioural variability can become costly. The pressure shifts towards conformity and the punishment of deviance. Just how profoundly rules can influence human behaviour was demonstrated once again in 2025 in a series of experiments by Simon Gächter, Lucas Molleman and Daniele Nosenzo involving over 14,000 participants: between 55 and 70 per cent complied with an arbitrary and materially costly rule even when acting alone and anonymously, without harming anyone through their own breach of the rule.

Leadership resolves another recurring problem in group life. When a collective decision needs to be made quickly, a hierarchy can reduce the cost of coordination. There is no need to imagine some sort of ‘authoritarianism gene’ – a category devoid of biological meaning: it is enough to recognise that leadership and followership can reduce uncertainty about who makes the decisions and how others will behave. Furthermore, if circumstances suggest a conflict with another group, the apparent effectiveness of a dominant leader – capable of imposing coordination and confronting the opponent – increases. In 2025, a study involving 5,008 people across 25 countries tested four independent predictions derived from this evolutionary hypothesis, finding overall that the presence or perception of intergroup conflict increases the preference for dominant leaders.

Under threat, therefore, several parameters of decision-making may change simultaneously. It becomes advantageous to react on the basis of less reliable signals; uncertainty makes social information more valuable; collective danger increases the returns from coordination and the cost of deviance; conflict between groups may increase the value attributed to dominant leadership. This is not a single ‘instinct’, nor is it a political programme encoded in our genes. These are different systems, produced by different selective pressures, which under certain conditions can drive behaviour in the same direction.

Natural selection, after all, does not produce universally correct decision-making procedures. It favours mechanisms that exploit the regularities of the environment in which they are selected and which, on average and under relevant conditions, increase the reproductive success of their bearers. If the relationship between the available signals and the consequences of different responses changes, a mechanism that was previously advantageous may begin to produce systematic errors. Behavioural ecology has long studied cases of this kind in animals exposed to rapidly changing environments, in which perceptual and decision-making systems shaped by a previous selective history may respond inadequately to new signals.

A predator attack, a raid by a rival group or a local resource crisis all present circumstances in which rapid decision-making, cohesion and a reduction in behavioural diversity can offer enormous advantages.

The European demographic crisis, immigration, industrial competition from China, the transformation of the labour market brought about by artificial intelligence, climate change, a pandemic or the funding of a pension system, however, belong to a different category of problems. The causes are widespread, interact with one another, produce effects on different time scales and respond to interventions through feedback loops that often generate unforeseen consequences. To tackle them, we need to keep multiple hypotheses open, gather independent information, tolerate dissent, experiment with competing solutions and wait long enough to measure their effects: precisely some of the steps that become more difficult to sustain politically when a situation is perceived as an urgent threat.

The growing complexity of the threats facing Western societies therefore calls for greater analytical rigour, whilst at the same time the perception of those very threats may encourage a decision-making approach geared towards acting swiftly, aligning with the group and rapidly reducing uncertainty. This paradox does not stem from some mysterious degeneration of contemporary cognitive abilities. It arises from the fact that a set of responses with an evolutionary logic can be triggered by problems whose solutions require very different behaviours.

Populist politics, almost by its very nature, has tools at its disposal capable of fuelling this dynamic. A complex phenomenon is portrayed as an immediate threat; a diffuse cause is attributed to the actions of identifiable actors; these are grouped together as an external or internal group to which the blame can be laid; dissent can be portrayed as weakness or disloyalty; finally, a leader promises to remove the obstacles that would prevent a swift response.

Such a portrayal produces something that serious politics can hardly promise with the same speed: a subjective reduction in uncertainty. A statement such as ‘the problem is immigrants’ has infinitely less explanatory power than an analysis of the demographic, economic and fiscal transformations of a European society; yet, in a matter of seconds, it identifies a cause, pinpoints a group to blame and suggests a course of action. The solution may be ineffective in addressing the problem itself, yet highly effective in creating the impression that the problem has become understandable and manageable.

The practical failure of a policy does not necessarily mean it loses its appeal. If part of its function is to demonstrate that the group is responding, that a boundary is being re-established or that someone is exercising command, this outcome can be perceived immediately, long before the tangible consequences become measurable. Furthermore, any failure may be interpreted as evidence that the threat is even more serious, that the enemies are more numerous or that the action has been thwarted, thereby providing fresh fuel for the demand for strength and coordination.

A self-sustaining cycle is thus formed. The threat increases the demand for simplification; simplification identifies an enemy; the enemy makes the threat more tangible; the heightened threat increases the value placed on cohesion and leadership. The discussion regarding the complexity of the causes may end up appearing suspect in itself, because those who raise doubts and qualifications seem to be holding the group back whilst danger looms.

There is therefore no need to imagine that millions of citizens in so-called liberal democracies have suddenly lost the ability to reason. Reacting swiftly to dangers, utilising information from others, coordinating with one’s group and modifying behaviour in accordance with the costs of error are all part of the normal repertoire of a social species. Their effectiveness, however, depends on the structure of the environment in which they are employed.

The scientific method, statistical analysis, cross-examination, the separation of powers and the institutional procedures that keep dissent alive long enough to allow hypotheses to be tested are all part of a much more recent history. These are cultural mechanisms that enable us to correct the errors to which biologically ancient decision-making systems may expose us when the world presents problems different from those that shaped their evolution.

Independent verification, specialist expertise and procedures that prevent a single individual from making an immediate decision come at a cost in terms of time, coordination and simplicity. During a crisis, it is precisely this cost that becomes particularly apparent and frustrating. It is also the moment when those procedures prove most valuable, because they protect a society from the risk of turning a perfectly understandable response to a threat into a collectively disastrous decision.

The difficulty with major modern crises lies in preserving the cognitive, informational and institutional conditions in which solutions can be sought, precisely at the very moment when the perception of danger makes speed, conformity and authoritarianism seem more appealing. It is then that we need reason most of all, and it is then that our evolutionary history makes it easier to choose something different.

The change is easily gauged by looking at who currently holds those positions. Robert F. Kennedy Jr. heads the Department of Health after describing the Covid vaccine as ‘the deadliest vaccine ever made’; Donald Trump withdrew the United States from the World Health Organisation and drastically reduced America’s involvement in global health, whilst Rand Paul continues to treat the origins of the pandemic as a potential criminal case in which Fauci should be held personally accountable. The recent Senate hearing, during which Fauci repeatedly invoked the Fifth Amendment, is part of the same dynamic. Six years on from the start of the Covid-19 pandemic, dissent over how certain decisions were made now coexists with a deeper transformation: the scientific uncertainty that accumulated during the crisis is increasingly being reinterpreted through the prism of personal responsibility and blame.

Fauci’s return has reignited this debate because his interview concerns not only the accusations levelled at him personally, but also the way in which scientific and political institutions made decisions whilst information about the pandemic was still incomplete.

A pandemic forces us to make decisions whilst our knowledge is incomplete, and Fauci’s interview brings this difficulty back into focus because many of the accusations levelled today at the handling of Covid treat decisions made when the data was still incomplete as evidence of incompetence or bad faith. In January 2020, crucial information on the transmission of the new coronavirus was lacking; there were no vaccines or specific treatments, and even estimates of the case fatality rate depended on denominators that were still uncertain. In the months that followed, knowledge accumulated regarding asymptomatic transmission, aerosols, the effectiveness of face masks, the duration of immunity and, finally, the ability of vaccines to prevent different forms of the disease and the emergence of new variants. In such a context, the quality of the scientific response also depended on how quickly recommendations were revised as the data changed.

Fauci emphasises this very point when he points out that science gathers the information available at a given moment and revises its conclusions as that information grows. This is a basic definition of the scientific method, yet it has been rendered politically explosive by our memory of the pandemic, in which a change to a recommendation is easily interpreted as proof that the previous one was fraudulently false, whilst an error in forecasting can become a sign of incompetence or bad faith. A serious reconstruction should instead establish what evidence was available when a decision was taken, what alternatives were realistically considered, and when new data should have altered the course of action; for only this chronology allows us to distinguish between a reasonable choice that turned out to be wrong and a choice that, even at the time, ignored sufficient information to have avoided it.

This is also the point on which Fauci ought to go into greater depth. When he says that, looking back, “I’m not sure we would change much”, he is condensing several years of the pandemic and decisions taken by individuals with very different levels of authority into a single statement. It would be far more useful to know which choices he still considers justified in the light of the information available at the time they were made, in which cases he believes that updates came too slowly, and where he acknowledges errors in the way risks and uncertainties were communicated to the public. Such a discussion would enable genuine scientific self-criticism, which is meaningful precisely when it succeeds in pinpointing the error and reconstructing the circumstances surrounding it without automatically turning it into a confession.

The story surrounding the origin of SARS-CoV-2 shows, however, just how difficult this distinction has become. Fauci continues to regard a laboratory accident as a possibility and notes that the available data do not allow it to be presented as the established origin of the virus, a position that essentially coincides with that reached by the WHO’s SAGO group, according to which the weight of the evidence favours a zoonotic spillover whilst the incompleteness of the information coming from China leaves other scenarios open. In a normal scientific controversy, the persistence of uncertainty should lead to calls for new data and a ranking of hypotheses proportionate to the available empirical support. In the American political debate, however, China’s lack of transparency has often been incorporated directly into the explanation, to the extent that the absence of the information needed to prove a laboratory escape is used as one of the reasons for considering it plausible or even probable.

Fauci thus ended up becoming the focal point for issues that actually belonged to different levels of the pandemic response. His public influence was enormous, and his constant media presence helped to identify him with the entire federal health system, whilst the power to close schools, businesses and public spaces lay largely with governors, local authorities and other bodies. Over the years, this distinction has become blurred in the collective memory, to the point where the caricature he himself presents in the interview – when he says: “I created the virus. I shut the schools. I did everything” – makes perfect sense. The quip works because it sums up a real process of retrospective concentration of responsibility, in which a particularly high-profile figure ends up becoming the symbolic focal point where decisions and failures – which are in fact much more widely dispersed – converge.

Once a judgement takes this form, the evidence gradually loses its ability to alter it. A document may be absorbed into the narrative that has already been constructed; a correction may become further confirmation of the suspicion; and anything that is not documented may be explained away by the hypothesis of a cover-up. This is the stage at which a scientific controversy ceases to be subject to genuine scrutiny, because every possible finding is reinterpreted in a way that is compatible with the initial conclusion. At that point, it becomes difficult even to learn from the actual mistakes made during the pandemic – mistakes which do exist and deserve to be studied – because a mistake that contributes to knowledge always requires the possibility of determining what might have prevented it.

Fauci’s reappearance comes as this transformation has now reached federal institutions. The Secretary of Health can describe Covid vaccines as exceptionally lethal despite the wealth of data accumulated on their safety, the Trump administration has withdrawn from the WHO and reduced the US presence in international surveillance networks, just as new outbreaks continue to demonstrate how crucial it is to recognise an outbreak quickly; and the still-necessary discussion of the mistakes made during the Covid pandemic is taking place within a system that is undermining some of the very tools through which those mistakes could be identified and corrected.

For this reason, a review of Fauci’s actions remains necessary, but it must not turn into a public trial in which politics and public opinion take on the role of a court called upon to pass judgement. Instead, the discussion must take place in accordance with scientific criteria: we must attribute to each individual what they actually decided, reconstruct the information available at the time that decision was taken, and distinguish between error and deception. Its aim is not to identify a culprit to be paraded as a scapegoat, but rather to understand what happened, which decisions were reasonable in the light of the data available at the time, and which might have been different, so that we may learn before the next emergency. The accountability of experts and the possibility for knowledge to evolve alongside the data are part of the same system of checks and balances: a public science in which every revision is retrospectively turned into a fault will end up defending its initial assertions even when the evidence has begun to contradict them.

The next epidemic will once again arrive with incomplete data and decisions to be made before all the consequences are known. The quality of the response will therefore also depend on the ability of institutions to quickly recognise what is not working and to correct it without every change being immediately interpreted as evidence of prior fault. After a pandemic in which we learnt almost everything whilst the crisis was already unfolding, it would be particularly dangerous to build a system in which maintaining a mistake becomes more convenient than correcting it.

Biodynamic agriculture provides a particularly clear example because it takes us beyond the realm of medicine, where illness and the need for treatment make the search for unfounded solutions easier to understand. Biodynamics stems from the lectures given by Rudolf Steiner in 1924 and retains practices that incorporate cosmic influences, preparations made by filling cattle horns with manure, and other procedures linked to its author’s anthroposophical conception; Nevertheless, in May 2021, the Senate debated a bill on organic farming in which biodynamics was explicitly mentioned, and Elena Cattaneo, a life senator and pharmacologist, had to table an amendment to remove that reference, which the Senate rejected. In the final text approved in 2022, the term was removed, whilst a general provision remained allowing methods that comply with European and national legislation to be treated as equivalent to organic farming. 

In January 2023, the 37th International Congress of the Association for Biodynamic Agriculture was held in the Adalberto Libera Aula Magna at Roma Tre University, under the patronage of, amongst others, the Ministry of the Environment, the Lazio Region and Roma Capitale; the programme featured speakers from academia and public institutions, as well as a keynote lecture on ‘Biodynamic identity, economics and social threefolding’, whilst the pre-conference sessions included presentations on biodynamic agriculture and anthroposophical medicine as ‘two daughters of Anthroposophy’ and on the role of intuition in scientific knowledge. Steiner’s theory did not need to change in order to attain a different social status, for it was sufficient simply to change the setting in which it was presented: within a university lecture theatre, amongst lecturers and representatives of public institutions, what outside that setting would first and foremost have required proof of its validity could now begin to draw on the very presence of the university as a mark of respectability. 

In October 2025, the mechanism reappeared in such an extreme form that it seemed to have been deliberately constructed to make it visible, when, on the initiative of Gian Marco Centinaio, Vice-President of the Senate for the Lega party, a conference was held in the Senate’s Nassirya Fallen Soldiers’ Hall dedicated to the so-called ‘Majorana Machine’, the device which, according to supporters of Rolando Pelizza, was allegedly designed in secret by Ettore Majorana and is said to be capable, amongst other things, of producing unlimited energy, annihilating matter, transforming common materials into gold and even rejuvenating living beings – without any of these claims ever having been subjected to independent experimental verification. The following January, Elena Cattaneo organised a CICAP conference in the same Senate entitled "La macchina di Majorana, una storia senza prove", specifically to reconstruct the events and clarify that the photographs, letters and videos presented by the supporters are no substitute for a verifiable experiment. 

This affair is almost a practical demonstration of the transfer of credibility that occurs when an institution lends its premises and its symbols. A member of the public does not have the means to personally verify particle physics, just as they cannot replicate a clinical trial before taking a medicine, and so they rely on reasonable indicators of reliability to guide them: a university, a research body, a ministry, Parliament. When a story about the transmutation of matter is brought before the Senate by an institutional body, that story inevitably gains something which its supporters can present to the outside world as endorsement, even when the accompanying statements make it clear that the Senate is not responsible for the opinions expressed.

Rino Liuzzi himself, known by his stage name ‘Jupiter’ and who works as an astrologer, chaired a conference organised by the Italian League for the Fight against Cancer (LILT) on 5 December 2024 at the Ministry of Health’s auditorium, dedicated to the current state of affairs and future prospects of the LILT; A month later, on 4 January 2025, he was invited onto Rai Radio 1’s programme “Caffè Europa”, whilst the future of space exploration was being discussed on the occasion of the European Space Agency’s fiftieth anniversary. The stars do not become capable of predicting events simply because an astrologer steps into a ministry or appears on a public service programme; rather, it becomes much more difficult to distinguish the astrologer from an expert when two institutions – which the public regards as gatekeepers of expertise – include him in their programmes. 

The university has an even more powerful impact, because here qualifications take the form of credits and degrees. In the 2025–2026 academic year, the Faculty of Medicine and Surgery at the University of Rome Tor Vergata continues to offer a one-year first-level master’s degree in ‘Natural Medicine and Naturopathic Sciences’, worth 60 credits, explicitly presented as a professional development opportunity; the programmes and course prospectuses associated with this course have included homeopathy, homotoxicology, Bach flower remedies and other disciplines of vastly differing scientific rigour. Studying homeopathy at university would be perfectly reasonable if it were treated as a subject within the history of medicine, sociology or critical analysis; however, a vocational Master’s programme housed within a Faculty of Medicine sends a different message, as it teaches students to use these practices and, as a result, awards university credits for them. 

The same ambiguity is inherent in the continuing professional development system for healthcare professionals. In 2017, the National Commission for Continuing Professional Development ruled that events dedicated to non-conventional medicine may only be accredited if they present technical and scientific content based on proof of efficacy and evidence-based medicine; however, it explicitly included homeopathic medicine, homotoxicology, traditional Chinese medicine, Ayurvedic medicine and anthroposophic medicine, as well as practices that raise different scientific issues such as phytotherapy and acupuncture. The clause on evidence-based medicine should resolve the contradiction, provided that the actual accreditation distinguishes between a course that teaches how to critically evaluate homeopathy and one that trains participants in prescribing homeopathic remedies; otherwise, what remains is primarily the visible administrative outcome, namely an activity that produces the same CME credits with which doctors certify their professional development.

Tuscany has turned this ambiguity into a structural health policy, stating on its official website that acupuncture, traditional Chinese medicine, homeopathy, phytotherapy and manual medicine “form part of the treatments covered by the regional health system”; as recently as April 2026, the Region listed four centres of excellence for complementary medicine, including a Homeopathy Clinic run by the North-West Tuscany Local Health Authority, and in the following May published issue 60 of its ‘Regional Bulletin on Complementary Medicine’, which featured, amongst other things, an article on the twenty years of experience of the homeopathic centre in Lucca.

The administrative category chosen by the Regional Authority downplays differences that are scientifically decisive, because phytotherapy involves pharmacologically active substances that can be studied using standard pharmacological methods, whilst homeopathy is based on the law of similars and on preparations obtained through dilutions and ‘potentisation’ – which the Region’s own page describes without clarifying to the reader the gap between this system and contemporary knowledge of chemistry and pharmacology. Members of the public find this description on the very same portal where they search for hospitals, healthcare provisions and services, with the result that the distinction between acknowledging the existence of a practice and attributing therapeutic legitimacy to it becomes increasingly difficult to discern. 

Bill S.1251, tabled by Senator Orfeo Mazzella of the Five Star Movement and still under consideration by the Senate, introduces this confusion directly into the wording of the legislation, as it refers to "pluralism in science", declares that it recognises the “therapeutic value of complementary and integrative therapies” and provides for these to be included within the services provided by the National Health Service, whilst subsequent articles regulate committees, professional registers and training courses. The therapeutic value of a practice falls within the category of empirical propositions, as it must be demonstrated by measuring a benefit against appropriate controls; Parliament may decide which services to fund and how to organise them, whilst a majority vote adds no data to that produced by clinical trials. 

In addition to the academic, administrative and political dimensions of credibility, there is a particularly insidious one, as it stems from the authority of the court and the role that scientific expertise plays in the proceedings. The judge must also rule on matters where the solution depends on knowledge they do not personally possess, and so they rely on technical consultants; if the expert report poorly selects the literature, confuses a temporal sequence with a causal relationship, or attaches disproportionate weight to minority studies, the scientific error may find its way into the judgement and immediately acquire a new authority. From that moment on, the process can become self-perpetuating, because the judgement is cited outside the courtroom as proof that a court has ‘recognised’ that link; this supposed proof reinforces the public credibility of the experts who had supported it, and the precedent may fuel new cases in which similar expert reports are presented.

In 2012, the Rimini Labour Court recognised a causal link between the MMR vaccination and autism in a child and ordered the Ministry of Health to pay the compensation provided for by law; In 2015, the Bologna Court of Appeal overturned the decision following a new expert report, deeming the studies used to support the link to be scientifically irrelevant; these included research by Andrew Wakefield, whose article on the link between the MMR vaccine and autism had already been retracted and whose scientific conduct had been judged to be seriously flawed. Meanwhile, the Rimini ruling had attracted considerable attention in the media and within anti-vaccine circles, because the claim that ‘a court has ruled that the vaccine causes autism’ carries far greater communicative weight than an epidemiological review explaining why no such link exists.

The same pattern has re-emerged with mobile phones. In 2017, the Court of Ivrea recognised an acoustic nerve neurinoma suffered by a Telecom employee who had made extensive use of a mobile phone for work purposes as an occupational disease, and in January 2020 the Turin Court of Appeal upheld the decision, arguing that there was a ‘scientific law of coverage’ capable of supporting the causal link; During the same period, a report on radio frequencies and cancer produced by the Italian National Institute of Health in collaboration with other scientific bodies presented a review of the evidence which did not confirm an increase in brain tumours attributable to mobile phone use, consistent with the findings of the large-scale Interphone study.

The issue here does not concern a worker’s right to claim compensation – which must be determined in accordance with the rules governing social security proceedings – but rather the transfer of a judicial conclusion from the realm of law to that of science. Civil proceedings may use evidential and causal criteria designed to resolve a dispute between specific parties; epidemiology, on the other hand, seeks to establish whether an exposure actually alters the risk within a population, through studies that must control for confounding factors, systematic errors and random fluctuations. When an individual court ruling is transformed into the statement ‘the courts have proven that mobile phones cause cancer’, two distinct systems of evaluation are conflated, and the authority of the judiciary ends up superseding that of scientific evidence.

The Xylella affair demonstrates just how destructive this cycle can become, even outside the field of medicine. In December 2015, the Lecce Public Prosecutor’s Office ordered the preventive seizure of the olive trees affected by the bacterium containment plan and entered ten people in the register of suspects, including researchers from the CNR, university lecturers and the special commissioner Giuseppe Silletti; the measure effectively halted the eradication operations provided for in the plan, whilst the prosecution’s case also called into question the scientific assumptions attributing a central role to Xylella fastidiosa in the dieback. In 2019, the investigation was closed with regard to the charges of spreading the disease, and in February 2026 the investigating magistrate in Bari also dismissed the final proceedings against the former head of the CNR, Donato Boscia, recognising that he had acted in accordance with European and national legislation. In the meantime, the disease had continued to spread throughout the region.

A judicial investigation inevitably possesses a power that a scientific article does not: it can seize materials, halt an activity, question people and turn a scientist into a suspect. Precisely for this reason, the entry of a scientifically weak reconstruction into the judicial system has consequences that exceed those of any pseudoscientific conference, because the error becomes capable of materially altering the reality on which it purports to pass judgement. If containment measures are halted during a plant epidemic, the time lost is not made up once the investigation is closed; if a researcher is portrayed for years as a possible culprit in the spread of the disease they are studying, a final acquittal does not automatically restore the public trust that has been lost.

Stamina took the relationship between the courts, politics and pseudoscience a step further. From 2012 onwards, several judges ordered the Spedali Civili hospital in Brescia to continue or administer the infusions requested by patients, whilst AIFA had already contested fundamental aspects of the production and characterisation of the cells; in 2013, Parliament intervened, allowing certain treatments to continue and paving the way for a publicly funded clinical trial. A procedure that had failed to produce the evidence normally required for an advanced therapy thus managed, in succession, to secure the endorsement of a public hospital, that of numerous court orders and, finally, that of the law; each of these was used to make it more difficult to return to the original question of what evidence actually existed.

This mechanism can be termed ‘epistemic recycling’, because a credential obtained for administrative, professional, judicial or political reasons is reintroduced into the debate as if it were scientific evidence. The registration of a product is cited as recognition of its efficacy; a master’s degree as academic recognition of the discipline it teaches; a court ruling as proof of causality; and a seat in Parliament as evidence that a theory at least deserves to be considered on a par with others. After a few such steps, the pseudoscientific practice has amassed a sufficiently substantial institutional track record to be used against those who still demand the data. The power of this procedure lies in the fact that every single act may be perfectly legitimate within its own scope, whilst the sum of these acts produces a result that no one seems to have formally decided upon.

The high level of education among those involved offers far less protection than we tend to assume, since a thorough knowledge of a discipline requires different skills from those needed to recognise a pseudoscientific construct when it has learnt to mimic the social structure of science. A doctor may be an excellent clinician yet be unable to properly assess literature that has been selected in a biased manner; a judge may apply the law impeccably yet rely entirely on their own adviser when it comes to understanding an epidemiological report; a professor may defend the presence of a baseless theory at their university by invoking academic freedom, without distinguishing between the freedom to study it and the credibility conferred by teaching it professionally. The fragmentation of responsibilities allows each person to deal correctly with their own part, whilst no one is any longer accountable for the epistemic quality of the final result.

Sometimes the consequences of this confusion are irreversible in the simplest sense of the word. Francesco Bonifazi was seven years old when, in May 2017, he died after a bacterial ear infection had been treated for days using only homeopathic remedies; the doctor, Massimiliano Mecozzi, was convicted of manslaughter at first instance and on appeal, and in January 2025 the Court of Cassation upheld the three-year sentence. In this case, homeopathy is neither an epistemological curiosity nor a matter of cultural freedom, because the replacement of an effective treatment with a pseudo-treatment led to the deterioration of a treatable illness, ultimately resulting in encephalitis and death. 

Misinformation about vaccines causes harm through a dynamic that is less easily attributable to individual propagandists but is observable at a population level: in the two-year period 2017–2018, Italy recorded 8,078 cases of measles and thirteen deaths, whilst the Italian National Institute of Health identified low vaccination coverage, which had built up over the years, as the main cause of the two epidemic waves; Almost 89 per cent of cases with a known vaccination status involved unvaccinated individuals. A ruling that erroneously attributes autism to the MMR vaccine inevitably enters the very same information environment in which families must decide whether to vaccinate, and its subsequent scientific refutation has far less capacity to circulate than the initial news story.

Death is the most immediately recognisable consequence of pseudoscience, whilst the wider damage begins when its practices infiltrate the institutions that a society relies on to determine which information is trustworthy. No individual citizen can personally verify the safety of a bridge, replicate a drug trial or sequence the genome of a bacterium before accepting a diagnosis; a complex society functions because it distributes knowledge and establishes institutions to which it entrusts the task of verifying it. If universities, public authorities, courts and Parliament cease to distinguish consistently enough between the social existence of a belief and its scientific validity, it is precisely this system of delegated trust that is compromised.

The consequence directly affects democracy, because a democracy can only withstand the conflict between interests and values as long as there remains sufficient common ground on which to establish what has happened and what the foreseeable effects of different decisions are. It is possible to discuss democratically whether to fund an effective treatment, how much to spend on controlling a plant disease, or what level of industrial risk to consider acceptable; but if the relationship between treatment and efficacy, between bacteria and disease, or between exposure and cancer becomes, in turn, a matter to be determined by parliamentary majorities, court rulings or political affiliations, the conflict ceases to concern what we wish to make of reality and begins to concern which reality each group has the right to proclaim.

For this reason, it would be extremely dangerous to respond by turning bad science into a criminal offence or by entrusting the state with the task of certifying which theories are scientifically correct. The opposite risk is already evident in the United States, where Donald Trump, the Republican president, signed Executive Order 14303, ‘Restoring Gold Standard Science’, in May 2025, requiring federal agencies to align their scientific activities with criteria defined by the executive; many of the principles listed in the text, such as reproducibility, transparency, falsifiability and the reporting of uncertainties, are unquestionably part of good scientific practice, whilst the institutionally sensitive aspect lies in conferring upon the political authorities the role of defining and enforcing from above the standard to which the science practised by federal agencies must conform.

Defending against pseudoscience therefore requires something more challenging and less spectacular than a ‘truth police’: universities, professional bodies, public administrations, the judiciary and politicians must take responsibility for the significance of the qualifications they award, whilst maintaining the distinction between the right to advocate any idea and the right to have that idea endorsed by a scientific or public institution. Academic freedom protects the study of homeopathy, but does not oblige a Faculty of Medicine to teach it as a therapy; political pluralism protects those who believe in biodynamics, without equating Steiner’s preparations with experimental agronomy on scientific grounds; access to justice guarantees everyone the right to seek compensation, without turning the court into the forum where it is determined, on behalf of the scientific community, whether a mobile phone causes cancer.

When, in 2023, the University of Catania cancelled an event entitled ‘The Reality of Homeopathy’, scheduled to take place in the Department of Pharmaceutical Sciences, the Vice-Chancellor, Francesco Priolo, did not forbid homeopaths from expressing their beliefs, nor did he decree which theory was true: he simply decided what content his department would endorse under the university’s name.

Pseudoscience can therefore be deadly, as happens when a pseudo-therapy replaces an effective treatment, but it is capable of even greater harm when it slowly poisons institutions, because from that moment on it begins to destroy the mechanism through which millions of people – who are, by necessity, unable to verify every specialist claim for themselves – decide whom to trust. If a citizen encounters biodynamics in universities and legislation, homeopathy in postgraduate courses and the National Health Service, a machine capable of producing gold in the chambers of the Senate, and a scientifically unfounded causal link within a court judgement, their difficulty in distinguishing between science and pseudoscience can no longer be explained solely by their ignorance. We have taught them that institutional credibility and the quality of evidence can be separated, and when this lesson is fully learnt, it is not just trust in science that is lost: the ability to distinguish between the authority derived from the verification of facts and that derived from the power to pronounce them is also lost. A democracy in which this distinction becomes permanently unclear remains formally capable of voting, legislating and judging, whilst progressively losing the ability to make decisions about reality itself.

When a lunar eclipse occurs

A lunar eclipse occurs when the Sun, Earth and Moon are aligned, or nearly so. The Earth lies between the Sun and the Moon and casts a cone of shadow into space, which may cover the Moon’s surface either entirely or only partially. This alignment occurs during the full moon phase, but not at every full moon, because the Moon’s orbit is tilted relative to that of the Earth. The portion immersed in the shadow does not disappear completely: some of the sunlight passes through the Earth’s atmosphere before reaching the Moon’s surface. The air scatters the blue components of the light in particular and allows the red and orange components to filter through more readily, following a mechanism similar to that which colours sunsets. The final hue can vary depending on the amount of dust and clouds present in the atmosphere: the greater the number of particles, the more intense the red may appear. This favours warm tones, ranging from copper to the deepest red, giving rise to the popular names ‘Red Moon’ or ‘Blood Moon’.

How and when to view it

The eclipse will be visible, at least in part, across large areas of the Americas, Europe and Africa. The best conditions for observing the entire event will be in the Americas, whilst in much of Western Europe the Moon will set before the eclipse ends. In Italy, the eclipse is expected to begin at around 4.40 am, whilst the peak, as mentioned earlier, should be reached at 6.12 am, with variations depending on geographical location. The event will be particularly well-suited for viewing in the central and southern regions, whilst in the northern regions, the Moon will set shortly after the peak. This means that the final phase will not be visible: the Moon will remain low on the western horizon, whilst the brightness of dawn will gradually increase. For this reason, it will be necessary to choose a spot with an unobstructed view to the west, free from buildings, trees or hills. Weather conditions will play a decisive role: clouds, mist or obstacles along the horizon could prevent you from observing the peak phase. The penumbral phase – during which only part of the Sun’s rays is blocked by the Earth and which, in this case, will begin during the night – will, however, be barely discernible without specialised equipment. Unlike a solar eclipse, observing a lunar eclipse does not require any filters or special protection: the phenomenon can be viewed with the naked eye.

No one had claimed that Rai intended to remove educational programmes from its archives or to stop producing them altogether. The issue was quite different: why shut down the television channel that broadcasts them? To reply that the content will continue to exist is to shift the focus of the question.

The main justification is that Rai Scuola needs to become "more digital" and bring content "to where the students are", making it available when needed via connected televisions, computers, tablets and smartphones. This argument would make sense if Rai Scuola were finally to launch online. It is regrettable that this is already the case. Today, Rai Scuola is available live on RaiPlay, alongside the other Rai channels, whilst its programmes are already available on demand; RaiPlay Learning, for its part, has been around for years.

What Rai has announced is therefore not the switch of Rai Scuola from linear to digital broadcasting. Digital broadcasting already exists and will be enhanced; what is being phased out is the linear channel. Currently, Rai Scuola is available both via the linear channel on digital terrestrial television and via RaiPlay and the web; from 1 October, the former will be discontinued, whilst the online service will be reorganised and enhanced. Calling this removal an ‘evolution in distribution and consumption methods’ does not make it an addition.

It is equally peculiar to argue that "educational content must be available when students need it, not just when it is broadcast". Certainly, and that is precisely why on-demand services already exist. No one is suggesting replacing it with the television schedule. The two forms of distribution are complementary: one allows you to search for a lesson when you need it, whilst the other also reaches those who have not logged into RaiPlay with the specific intention of looking for it. The advantage of the former justifies its expansion; it provides no reason to eliminate the latter.

Even the phrase "more accessible" in the headline is a claim, not a proven fact. For some users, the new service may certainly become better and easier to use; for viewers who currently access Rai Scuola via their television, one means of access is simply being removed. To determine whether more people will ultimately be reached, one would need to compare the audience gained through digital enhancement with that lost by abandoning the linear channel. Rai does not present this comparison.

You should be well aware of the importance of this. In its 2024 annual report, Rai notes that, following the switch to DVB-T2 and the consequent redistribution of channels across the multiplexes, for those directly affected (Rai Scuola, Rai Storia and Rai Radio2 Visual), the decline in net Auditel viewership figures was close to 50 per cent. The programmes were still there; however, due to the changed conditions of access, a huge proportion of the audience could no longer reach them. This is proof, provided by the company itself, that the availability of content and the ability to reach the audience are not one and the same thing.

The press release then attempts to use viewing figures against the channel, citing the 0.14 per cent audience share in 2024, the 0.13 per cent in 2025 and the decline in the early months of 2026, concluding that these figures demonstrate "the limitations of linear television as a means of reaching the school-age audience broadly and systematically". Here, the problem is no longer merely the presentation: it is the reasoning. A share calculated across the entire television audience indicates what proportion of that audience is watching Rai Scuola; it does not measure the ability to reach the "school-age audience". To make the latter claim, data on penetration amongst pupils and teachers would be required, along with a comparison with the figures achieved via RaiPlay. The press release does not provide these.

It also omits a key fact necessary for interpreting those figures: the 0.13–0.14 per cent for 2023–2025 represents one of the highest levels achieved by Rai Scuola in its recent history. The channel is therefore being axed after a three-year period in which it had consolidated viewing figures well above those of previous years. In 2026, there is a decline, and this should be noted; however, it is not enough to retroactively transform the previous growth into proof of the failure of linear broadcasting.

And that 0.14 per cent does not correspond to a non-existent audience: in February 2024, for example, a 0.14 per cent share corresponded to an average of 13,322 viewers, i.e. those present on average every minute of the day. Daily net contacts are necessarily a much higher figure, in the hundreds of thousands, but Rai does not provide these figures in its press release. This is not the audience of a general-interest broadcaster, and no one could claim otherwise; it is the audience of a highly specialised educational channel – an audience that is an asset to be carefully preserved rather than squandered. Before claiming that this audience is better reached elsewhere, it would be useful to present the data to back up that claim.

Finally, there remains the point that renders the entire justification based on "moving beyond the linear model" particularly fragile. From 1 October, channel 57 will not be taken off air. It will be occupied by a new linear channel, dedicated to Italy. Rai presents it as a programme line-up centred on landscapes, villages, traditions, ‘Made in Italy’, food and wine, tourism, folklore and national events; it will also launch with a ‘re-run programme’, to which reruns, new productions and live broadcasts are expected to be added from the following quarter.

This settles the technical issue. Rai does not consider the linear thematic channel model to be obsolete, as it is launching a new one on the very same day that it is discontinuing Rai Scuola. It considers the linear broadcasting of educational content to be obsolete, whilst it considers the coverage of local communities to be worthy of linear broadcasting.

It may be an editorial decision. It may depend on the capacity available on the multiplexes and therefore on the need to decide which channel to allocate it to. It may even be a decision that Rai considers beneficial in terms of audience reach. However, none of these possibilities matches the explanation given in the press release. If there is only room for one channel and Rai decides to allocate it to Italiana, the issue concerns the priority given to two different functions of the public service.

This is why the final reference to ‘distracted observers’ – who are said to have missed a logic that is ‘so’ obvious – is particularly unfortunate. To a distracted observer, it may indeed seem as though an old television channel is finally being transformed into a modern digital platform. One need only look at what Rai already offers today to discover that the digital platform already exists, that Rai Scuola is already available on RaiPlay, and that the linear service – which is said to have become unsuitable for schools – is simultaneously considered perfectly suited to a new regional channel.

The question therefore remains exactly the same as before. If the new RaiPlay Scuola e Learning is set to improve the educational offering, why must we also give up the audience that Rai Scuola currently reaches via digital terrestrial television in order to improve it? And if the viewing figures really do demonstrate the "limitations of linear television", why is that very same linear television service immediately being handed over to Italiana?

The press release explains why Rai wishes to invest in educational digital content – a decision to which it is difficult to raise objections, and an idea that has, moreover, been discussed for some time; but this does not explain why, in order to do so, it must take Rai Scuola off Channel 57.

Writing unpersuasive press releases that grasp at straws through petulant responses serves no purpose whatsoever.

For many years, Rai Scuola had minuscule viewing figures. Between 2015 and 2021, its average audience share generally remained between 0.01 and 0.03 per cent, then in 2022 a growth trend began which, by 2023, had raised the annual average to 0.13 per cent and to 0.14 per cent in 2024, whilst in 2025 the channel maintained 0.13 per cent. The change introduced by Auditel to the way audience share is calculated in May 2022 calls for caution when comparing figures with previous years, but it does not explain the further growth seen in 2023 – when the new method was already in force – and, above all, it does not negate the fact that, for three years, Rai Scuola maintained viewing figures far higher than those seen in its previous history. In 2026, a decline was observed, with figures hovering around 0.10 per cent in the early months, whilst by June the audience share had returned to 0.14.

Taken out of context, these figures make it all too easy to dismiss Rai Scuola as a channel that nobody watches. A 0.14 per cent audience share corresponds to an average of around 12,000–13,000 people watching the channel every minute of the day. As viewers come and go, the number of different individuals reached each day is much higher: Rai’s historical data show that as early as 2016, when Rai Scuola’s viewing figures were significantly lower, the channel reached 176,000 people a day. Based on the 2024 viewing figures, the daily audience can reasonably be estimated at several hundred thousand and, depending on the average viewing time, even around one million.

Rai also has another metric, perhaps even more relevant to the mission of a public service broadcaster. In the Qualitel 2024 survey, Rai Scuola achieved an overall satisfaction rating of 8.0 for the television channel, ranking second amongst the thematic channels behind Rai Storia; for its digital offering, it scored 8.1, on a par with RaiPlay and at the top of the digital platforms considered. The same survey awarded the cultural and educational programmes on the general-interest channels the highest average score, 8.2. These figures were produced by Rai itself, based on a survey of 25,000 people, and describe an audience that places a high value on precisely the kind of content that the public service has decided to sacrifice (Rai Press Office).

One might understand this decision had a public service function deemed even more urgent been planned in place of Rai Scuola. However, the press release in which Rai presents the new channel points in a different direction. “Italiana” stems from the “well-established success” of regional reporting and the experience of “Linea Verde”; it will launch with a selection of programmes that have already been broadcast and, from the following quarter, will add re-edited programmes, new productions and live coverage of major national events relevant to the channel’s themes. The stated aim is to explore Italian identity through the regions (RAI Press Office).

The problem lies precisely in the choice of priorities that emerges from this substitution. Italy is already one of the subjects most frequently covered by public television, through programmes dedicated to villages, cooking, agriculture, tourism and local produce, whilst Rai itself cites "Linea Verde" as the inspiration for the new project. Schools have far less airtime, in which a maths lesson, a physics experiment, a history of literature or a biology lesson can be presented without having to compete with the demands of general entertainment. The Rai Scuola website offers thousands of lessons organised into more than thirty subjects and over eighty sub-categories, alongside humanities courses, science workshops and in-depth features (Rai Scuola).

It is difficult to see this transformation as an expansion of the public service, because a function that is difficult to replace is being swapped for one of which there is already an abundance on television. Public service television certainly has a duty to portray the country and can do so through its culinary traditions and local festivals; however, when it eliminates its own educational channel in order to make room for this portrayal, it is establishing a very specific cultural hierarchy. Subject-specific knowledge loses a dedicated platform, whilst the portrayal of Italy through its regions, local specialities and events gains an entire network.

The issue becomes even more serious when we consider what is happening in Italian schools. For years, national and international surveys have documented difficulties with basic skills and significant regional disparities; at the same time, the public dissemination of scientific knowledge must contend with an information environment in which unregulated content can reach millions of people. In such a situation, it would be reasonable to ask how to strengthen a public infrastructure dedicated to education, how to link it more effectively to the real-world school system, and how to transform the growth in viewership seen in recent years into an even wider audience. Rai has chosen to shut down that television infrastructure.

Naturally, Rai Scuola’s content will continue to be available online, and part of its archive may be preserved on RaiPlay. However, a digital library serves a different purpose to a linear channel: it showcases a range of programmes, allows for chance encounters with content and, above all, affirms – through its presence in the public sphere – that education deserves a recognisable place within the national television system. Transferring an archive to the web preserves programmes; closing the channel eliminates that presence.

Finally, there is one fact that should make this decision particularly embarrassing: Rai Scuola is being axed at a time when its viewing figures in recent years are amongst the best in its history and Rai’s own quality surveys show very high levels of satisfaction. It is therefore not the public who have deemed Rai Scuola to be unnecessary. It is Rai itself that has decided that this slot can be put to better use for a different vision of public service.

From 1 October, instead of a channel that brings schools to television, we will have one built around the narrative of regional identity, with programmes initially drawn from daytime archives and, subsequently, live coverage of national events. These will certainly include highly commendable cultural initiatives, alongside that world of fairs, local produce and food festivals which has now become a national television genre. It is a choice that at least has the merit of clarity: in a country where it would be difficult to argue that there is an excess of education, the public service broadcaster has found a slot on television to take away from schools and dedicate to the local area. From schools to sausage festivals, on the same channel.

Failing to get vaccinated when a vaccination is recommended, skipping cancer screening, smoking, abusing alcohol, and leading a completely sedentary lifestyle: these are behaviours that can increase the risk of disease and, on a collective scale, add to the burden of chronic conditions on the healthcare system. The World Health Organisation identifies tobacco, harmful use of alcohol and physical inactivity as some of the main modifiable risk factors for cardiovascular diseases, cancers, chronic respiratory diseases and diabetes. It emphasises that preventing these conditions not only saves lives but also reduces economic costs and the pressure on healthcare systems.

Even something as seemingly simple as being more active has enormous consequences. The WHO estimates that, if physical inactivity is not reduced, its cost to global healthcare systems could reach around 300 billion dollars between 2020 and 2030.

This should be one of the cornerstones of any serious policy: prevention is not a luxury; it is one of the most sensible forms of public investment.

Yet prevention is often the first casualty of health illiteracy. Cancer is only tackled once it has already developed, rather than attempting to detect it early through appropriate screening programmes. Vaccination is turned into an ideological issue rather than being assessed on the basis of the available evidence. Smoking is regarded as merely a personal habit, whilst overlooking the fact that its prevalence has enormous health and economic consequences. Alcohol is often excluded from discussions about risks because it is culturally accepted.

The issue, of course, is not to place the blame on individual citizens. People do not choose their health status in a vacuum: factors such as income, education, the environment in which they live, and the practical possibility of accessing treatment and preventative care all play a part, as do commercial and social pressures. This is precisely why the role of politics is crucial: to create the conditions that make healthy behaviours easier and more accessible, rather than merely criticising those who are unable to adopt them.

Then there is a second, even more worrying level: scientific illiteracy.

We saw this during and after the pandemic, when part of the public debate turned scientists, epidemiologists and scientific institutions into political targets. A prime example is that of Anthony Fauci. One can certainly debate his decisions, the strategies adopted during the pandemic, his public communications and any mistakes he may have made: science is not infallible, and no scientist should be exempt from critical scrutiny.

But it is one thing to offer evidence-based criticism; it is quite another to turn a scientist into a political enemy simply because his conclusions have become inconvenient.

Fauci is not simply ‘a TV doctor’. He served as Director of the National Institute of Allergy and Infectious Diseases from 1984 to 2022, led research programmes on HIV/AIDS, tuberculosis, malaria, Ebola, Zika and other infectious diseases, and is a member of the National Academy of Sciences and the National Academy of Medicine. He has also been awarded the Presidential Medal of Freedom.

This does not mean that Fauci has always been right. It means that his scientific track record should demand a serious approach to judgement – neither adulation nor demonisation. Science is not defended by turning scientists into saints; it is defended by demanding evidence, method, transparency and the ability to correct errors. And this is where the issue becomes political.

A democracy needs informed citizens. A citizen who understands the difference between correlation and causation, who knows what relative risk means, who understands why screening programmes exist, who can critically evaluate scientific research and who can distinguish an expert from a charlatan is much harder to manipulate.

On the contrary, a fearful, misinformed electorate lacking the tools to assess the evidence is extraordinarily vulnerable to propaganda.

It is therefore legitimate to question a policy that chooses to fuel mistrust of science rather than invest in scientific and health education. And it would be wrong to turn this criticism into a blanket accusation against "the right" as a whole: there are different positions and perspectives within all political camps. But it is equally legitimate to speak out when, particularly in certain sections of the political and populist right, mistrust of experts, scientific institutions and public health is used as a tool for electoral mobilisation. Because governing an informed public is more difficult.

We must explain the data. We must acknowledge mistakes. We must account for our decisions. We must consult with independent experts. We must invest in education and prevention. We must tell the public that certain choices have consequences not only for individuals, but for the community as a whole.

It is much easier, however, to construct an enemy: the ‘regime-aligned’ doctor, the scientist ‘enslaved by Big Pharma’, the ‘corrupt’ researcher, the ‘imposed’ vaccine, the healthcare institution ‘that wants to control us’. It is a politically effective narrative because it transforms complex issues into simplistic stories, with good guys and bad guys, victims and persecutors. But reality does not work like that. A public health system cannot survive on hospitals, A&E departments, expensive medicines and increasingly complex treatments alone. It needs citizens who are empowered to prevent diseases, recognise them at an early stage and make decisions based on the best available evidence.

The alternative is to continue spending enormous resources on treating conditions that, at least in part, we could have prevented. The WHO emphasises precisely that early intervention in non-communicable diseases can reduce the need for more expensive treatments and represent economically beneficial investments. True freedom, then, is not freedom from scientific evidence. It is the freedom to know.

And a society in which scientific knowledge is devalued, prevention is mocked and experts are automatically treated as enemies is not a freer society. It is a more fragile, more costly and, above all, more easily manipulated society.

Defending public healthcare also means defending scientific culture. Because whenever ignorance about health takes precedence over evidence, the bill eventually comes due: in people’s lives, in hospitals and in the state’s coffers.

Matteo Bassetti is a full professor of Infectious Diseases at the University of Genoa

For each patient, up to 34 tumour neoantigens are selected and an mRNA encoding them is synthesised. The immune system is thus trained to recognise residual tumour cells carrying those mutations. In phase 3, the combination improved both recurrence-free survival and distant metastasis-free survival. Merck and Moderna describe the benefit as clinically significant. The full data are not yet available, so the magnitude of the effect remains to be determined, but the signal already observed in Phase 2 has withstood the test of a large confirmatory study.

The result makes the short-sightedness of the MAHAs and the Trump administration even more apparent: divesting from a platform also means betting against applications that the platform has not yet had time to demonstrate. On 5 August 2025, the US Department of Health announced the federal withdrawal from twenty-two investments in the development of mRNA vaccines, totalling nearly $500 million. The programmes focused primarily on respiratory infections and pandemic preparedness. Robert F. Kennedy Jr., the chief liar at the Department of Health under the Trump administration, explained that those platforms had shown limitations in efficacy and that resources would be redirected towards technologies deemed safer.

That decision came after years in which mRNA had become embroiled in the controversy surrounding Covid vaccines, to the point where it had become a rallying cry for a section of the American right. Ron DeSantis’s Florida had paved the way. In January 2024, Surgeon General Joseph Ladapo had called for its discontinuation for the entire population, citing the risk that residual DNA fragments might integrate into the genome of those vaccinated; the FDA had responded that such a mechanism was biologically implausible and that the data showed no signs of genotoxicity. In 2025, Kennedy brought some of that same hostility to Washington.

The controversy had by now moved beyond Covid vaccines. In 2025, in Minnesota, an ‘mRNA Bioweapons Prohibition Act’ proposed classifying all products containing mRNA or modified mRNA as ‘weapons of mass destruction’. In 2026, a similar bill was tabled in Tennessee. Neither became law, but both illustrate the extent to which a field of molecular biology has been drawn into a political battle.

The federal measure concerned a specific aspect of mRNA applications, but a technological platform thrives on the scientific expertise and production capacity that accumulate as it is used. Reducing public investment restricts the scope within which that expertise is maintained and transferred to new applications. It is only private investment that has prevented the withdrawal of public funding from becoming a constraint on the technology as a whole.

There is also a second effect of US policies, which affects us directly. The ‘most-favoured-nation’ clause on drug prices aims to align US prices with the lowest prices charged in other developed countries. It may therefore become a rational decision for a company to delay launches in countries that negotiate lower prices, such as Italy, or to demand less favourable terms from public healthcare systems. Merck is among the companies that have entered into the Trump administration’s MFN agreements. We do not yet know whether Intismeran will be included, but the mechanism is clear: a low price negotiated in Europe can reduce the value of the US market and make it less profitable to bring the drug to us quickly.

The cost of this policy may therefore be borne by Italian patients. mRNA research risks being undermined in US public research and, when it leads to new medicines, may face barriers that delay access to them in Europe.

Thank you, really, you chief liar.

Climate change is significant for our well-being, but economic growth and the advancement of scientific knowledge have been incomparably more so. Of course, past performance is not indicative of future results and, as the experts tell us, change is a non-linear phenomenon. On the other hand, the same scientists tell us that the worst-case scenarios which, until recently (and still today for some), were presented as the outcome of ‘business as usual’ are now highly unlikely. As time goes on, reality is drifting further and further away from both the ‘deniers’ and the ‘doomsayers’.

However confident we may be in our ability to adapt, putting all our eggs in this basket is not the best option. We also need to reduce emissions. And, as Bucci writes, the best course of action in the medium to long term is to make decarbonised energy more cost-effective than that produced from fossil fuels and to lower the costs of CO₂ capture. Investing more in basic research is wise, whilst recognising that the greatest benefits will be for those who come after us and who, in any case, will inherit a far better world than the one we received.

Many of the policies adopted by the EU, however, are ill-advised because, amidst a tangle of increasingly detailed rules, they ignore the cost implications and, as a result, make the necessary process of building consensus far more problematic.

Putting a price on carbon, whilst channelling the revenue to the poorest people and/or using it to reduce emissions at much lower cost outside the EU, would be the most efficient and fairest policy. We can understand this clearly by looking at petrol. Bucci writes that: “One litre of petrol not burnt avoids the emissions associated with that litre.” This, however, is only one side of the coin. The other, which almost always goes unnoticed, is that for every litre of petrol not burnt, the state loses around one euro, which equates to €400 for every tonne of CO₂ not emitted. It is a terrible deal because, for a fraction of that amount (without even taking into account other forms of car taxation), one could eliminate an identical tonne of carbon dioxide, thereby making the transport sector carbon-neutral with immediate effect. Equally, the substantial resources that the EU intends to allocate to rail investments – which will have no appreciable effect on decarbonisation – are a poor investment. Fewer railway sleepers, fewer (and better) regulations, and more research is the formula to follow.