“They laughed at Galileo, so my controversial theory must be right!” This is the Galileo gambit, a logical fallacy that surfaced repeatedly in Europe during COVID-19, notably around French professor Didier Raoult and his promotion of hydroxychloroquine.

As Carl Sagan countered in Broca’s Brain:1 “… the fact that some geniuses were laughed at does not imply that all who are laughed at are geniuses. They laughed at Columbus, they laughed at Fulton, they laughed at the Wright brothers. But they also laughed at Bozo the Clown.”

Raoult’s Case

Professor Raoult, one of France’s most prominent and most-cited infectious-disease researchers, sidestepped standard randomised controlled trials for hydroxychloroquine and dismissed his international critics as a dogmatic, out-of-touch establishment.

He was seen by some as a persecuted maverick. He was described as a frontline doctor, focused on the immediate care of his patients at his institute in Marseille, in contrast to the bureaucrats and the Scientific Council, whom he deemed disconnected from clinical reality.

A prominent rhetoric among his defenders claims that the pharmaceutical industry targeted him. Because his hydroxychloroquine-based treatment was old and in the public domain, it generated no profit for major pharmaceutical companies. According to this theory, the criticisms directed at him were aimed solely at protecting the massive profits associated with new treatments and vaccines.

The multiple reprimands from the scientific community, the retraction of more than twenty of his studies (with expressions of concern issued for hundreds more on ethical and methodological grounds), and his two-year ban from practising medicine (handed down on appeal in October 2024 and effective from February 2025) are not viewed by his audience as standard scientific regulatory processes. Rather, they are experienced as a witch hunt orchestrated by the state and the media to silence him.

But the gambit relies on survivorship bias: it hides the far more numerous dissenters who were ridiculed and plainly just wrong. It smuggles in a claim it never proves: that institutional rejection is itself a mark of truth. It isn’t.

The mavericks who were dismissed and mistaken don’t get essays written about them. The survivors do the persuading. Raoult got the essays without the vindication.

What actually vindicates a maverick

What vindicated Galileo, or mavericks like Ignaz Semmelweis and Barry Marshall, was never the pushback itself. It was new evidence, a plausible mechanism, or a measurement technology that finally settled the question.

In the 19th century, Hungarian physician Ignaz Semmelweis was baffled by a paradox: mothers who gave birth on the dirty streets were inexplicably safer from puerperal fever than those in his maternity clinic. Determined to solve this, he traced the source of the infection, concluding that doctors were carrying microscopic “cadaverous particles” directly from the autopsy tables to women in labour. Semmelweis ordered a simple intervention: washing hands with chlorinated lime before patient examinations. The results were staggering, driving the clinic’s mortality rate down by about 90%.

Yet, rather than being celebrated, Semmelweis was ostracised. The medical establishment, entrenched in the ancient belief that illnesses stemmed from an imbalance of the body’s “four humours,” balked at his ideas. Prominent doctors were deeply insulted by the very suggestion that their hands were dirty. Tragic and ahead of his time, Semmelweis was only vindicated posthumously by the germ theory subsequently developed by Pasteur, Lister, and Koch, which proved why the doctor’s handwashing mandate had saved so many lives.2

Even more recently, in the 1980s, Barry Marshall, an Australian professor of clinical microbiology and the 2005 Nobel laureate, challenged medical doctrine by showing that a bacterium, Helicobacter pylori, plays a major role in causing peptic ulcers, when the majority thought they were mainly caused by stress, too much acid, or spicy food. Many scientists outright dismissed his claims, believing that no microorganism could survive the stomach’s acidic environment. When early attempts to infect animals like pigs failed, Marshall resorted to a radical (arguably crazy) self-experiment in 1984: he drank a broth full of live H. pylori cultures.

He expected to develop a slow, asymptomatic infection over the course of about a year. Instead, within days he fell ill, and a biopsy about a week later confirmed severe gastritis (stomach inflammation), with H. pylori colonising his stomach lining. This was the crucial step. It proved that the bacterium could colonise a healthy stomach and directly trigger the inflammation that degrades the protective mucus layer. That process that, left unchecked, lets ordinary stomach acid burn into the underlying tissue and form an ulcer. The self-experiment proved the bacterium was a genuine pathogen; the full link to ulcers, and the cure, were then nailed down by diagnostic tests and eradication trials, which showed that clearing H. pylori with antibiotics healed ulcers and stopped them recurring. It was that body of evidence, not the drama of the experiment, that transformed peptic ulcer disease from a lifelong condition into a curable bacterial infection.

In conclusion, Marshall was vindicated by diagnostic tests and eradication trials; Semmelweis, posthumously, by germ theory. The persecution or the ridicule was incidental. The evidence was decisive.

Why the gambit works

The gambit is not only a logical error. It is cognitively optimised to bypass evaluation which is precisely why “free speech” in matters of health is not as harmless as it sounds. Several well-replicated findings explain why people are moved more by a story about a persecuted genius than by a meta-analysis:

  • Narrative beats data. Absorption into a story — what psychologists call transportation — reduces counter-arguing and increases belief in story-consistent claims.3 A lone hero defying the establishment is a far better story than a confidence interval. Early in the COVID-19 epidemic, Raoult was appointed to France’s newly created scientific council on the virus but quickly stepped away and set himself against the government’s recommendations.
  • The identifiable victim outweighs the statistic. People feel for one named, vivid individual and go numb to aggregate numbers.4 Human empathy is asymmetrical: a single, named person facing a vivid struggle moves us, yet massive aggregate statistics leave us cold. One tragedy is a heartbreak; a million is just a math problem. The persecuted maverick is identifiable; the patients quietly saved by consensus medicine are a statistic.
  • Repetition manufactures truth. A claim feels truer simply for being repeated — even when the person already knows better.5 Social feeds are repetition machines.
  • Falsehood travels faster. In another study — roughly 126,000 stories on Twitter (now X), tweeted by about three million people more than 4.5 million times — false stories were about 70% more likely to be reshared than true ones, and spread faster and deeper. And it was humans, not bots, who drove the gap.6 Raoult’s hydroxychloroquine claims followed the same path: they crossed the Atlantic and fed directly into US President Trump’s endorsement of the ineffective treatment.

The cost is not abstract

Put those mechanisms together at scale in a health emergency, and the result is measurable.

The hydroxychloroquine episode is instructive, and it shows that science can, sometimes, be messy. The “evidence” cut in both directions: the most influential paper against the drug, published in The Lancet, was retracted within weeks when its data source, Surgisphere, would not release its underlying data for an independent audit.7 Meanwhile, the chloroquine-and-ivermectin hype produced hard casualties. U.S. poison-control calls for ivermectin rose to roughly five times their pre-pandemic baseline by mid-2021;8 and an Arizona couple ingested aquarium-cleaner chloroquine phosphate after televised endorsements of chloroquine, killing one of them.9 More broadly, U.S. poison centres logged a surge in cleaner and disinfectant exposures in early 2020 as pandemic anxiety took hold.10

The CDC’s analysts were careful: “Although a causal association cannot be demonstrated, the timing of these reported exposures corresponded to increased media coverage.” That caveat is worth keeping in mind — but, in my own opinion, the bodies were real.

The pattern generalises well beyond one drug:

  • Cancer. Patients with curable cancers who chose alternative medicine instead of conventional treatment were two and a half times more likely to die.11 A companion study found the excess mortality was driven specifically by refusal of effective treatment, not by the supplements themselves.12
  • HIV/AIDS. South Africa’s state AIDS denialism under President Thabo Mbeki, which delayed the antiretroviral rollout, is estimated to have caused more than 330,000 preventable deaths between 2000 and 2005.13 These are modelling estimates, not direct counts. Note that the persuasion vector here was not a salesman; it was a head of state, inside a democracy with a free press. The marketplace of ideas did not self-correct.
  • COVID-19 vaccines. Roughly 232,000 U.S. adult deaths between mid-2021 and September 2022 are estimated to have been preventable by vaccination.14 In a randomised experiment, exposure to recent online misinformation cut firm intent to vaccinate by about six percentage points (6.2 in the UK, 6.4 in the US), and “scientific-sounding” misinformation was more persuasive than overtly fringe content.15
  • Measles. As MMR uptake fell (England’s first-dose coverage at 24 months is now at its lowest since 2009–10) measles returned on both sides of the Atlantic. England recorded 2,911 confirmed cases in 2024, its most in over two decades, and a child died in 2025; the United States recorded 2,288 confirmed cases in 2025, its worst year in more than three decades, including three deaths, almost entirely among the unvaccinated.16

What makes consensus trustworthy

So what makes scientific consensus worth trusting? Not a single rigid “scientific method,” but a social process: claims vetted, contested, and revised by a diverse expert community. Consensus does not guarantee truth. Surgisphere is a reminder that the system self-corrects precisely because it is fallible. But it is a reliable indicator that a claim has survived collective scrutiny. It is the best current explanation for a phenomenon given the weight of evidence and rigorous peer review. So yes, it shifts; yes, it could be overturned tomorrow. But it remains the best tool available.

The procedural line

This is also what separates legitimate dissent from the gambit. Real dissent engages the criticism and shares the evidential standards of its field; it is not validated by the mere claim of suppression. The defensible test for scientific liberty is procedural: was a claim rejected after a fair hearing of its evidence, or arbitrarily dismissed? In public health, robust evidence, not authority and not perceived martyrdom, must lead.

Free speech, consensus, and the honest limits

None of this puts “free speech” and “scientific consensus” in genuine tension. You can publish, argue, and dissent freely. Consensus is just the standing record of what survived scrutiny: a verdict, provisional, not a gag order on dissenting ideas.

But “not in tension” is a claim about rights, not about consequences. The evidence above is exactly what the classic marketplace-of-ideas defence assumes away: that truth will somehow win, out-competing falsehood in open debate. In health, that self-correction often fails, because persuasion so often defeats it.

The tempting conclusion, “so let’s censor it”, does not survive contact with the evidence either. Where the line around “misinformation” falls is unstable; suppression manufactures martyrs and feeds the very gambit it means to stop; and the soft alternatives are weak.

Take accuracy nudges. They consist of prompting people to pause and consider whether a headline is true before they share it. Those nudges show promising effects in controlled studies, but the effect is small, and a preregistered replication of the key COVID-era result initially failed, recovering only a weak signal once the samples were pooled. Whether such nudges achieve much at all in the realistic churn of a social-media feed remains genuinely contested.17

The danger is real and measurable, and we do not yet have a scalable, well-evidenced cure. What that leaves is not silence and not censorship, but the procedural discipline above. Free speech should be preserved. What no one is owed is the presumption that having been laughed at is the same as being right.

A live test: Europe’s Digital Services Act

The cleanest current test of where that line should fall is unfolding in Europe. The Digital Services Act (Regulation (EU) 2022/2065, in force since 2022 and fully applicable since February 2024) is not, despite the shorthand, a “censorship law”: it sets tiered transparency duties on intermediaries, keeps the ban on general monitoring, and leaves the definition of illegal to national law. Its sharpest obligations fall only on the “very large” platforms and search engines, those above 45 million monthly EU users, which must submit to audits and open their data to outside scrutiny.18

Two of its mechanisms I support without hesitation, because they generate evidence rather than verdicts. Algorithmic transparency (Articles 27 and 38) forces platforms to disclose the main parameters of their recommender systems and to offer at least one feed not built on profiling.19 Researcher data access (Article 40), whose operating rules and one-stop portal went live in 2025, lets vetted researchers pull platforms’ internal data to study systemic risks and test whether mitigation works.20 This is the self-correcting machinery this essay defends: it decides nothing about truth, only lets outsiders measure amplification and check the platforms’ work. The new vetted-researcher channel is early and friction-heavy, though. A regulator in Ireland rejected one of the first-wave requests (a Dutch team seeking TikTok election-monetisation data) in February 2026 for failing five of the seven application requirements.21

I am warier of the other half: the duty to mitigate “disinformation” as a systemic risk. Articles 34–35 oblige large platforms to reduce “actual or foreseeable negative effects on civic discourse and electoral processes,” backed by a Code of Conduct on Disinformation — voluntary, but a compliance benchmark since July 2025.22 The aims are decent; the terms are elastic — and an authority empowered to define a “negative effect on civic discourse” sits one step from arbitrating truth. This essay is its own warning: Semmelweis was consensus-wrong, and Mbeki shows a state can be the lethal misinformation vector inside a free press. Machinery built to suppress the next deadly falsehood can, in other hands, suppress the next unpopular truth. In fairness, the strongest legal reading cuts the other way: the codes are voluntary, disinformation is never made illegal, and the Commission’s “red line” is that it cannot force platforms to delete lawful speech, only to label or add friction.23

Moreover, enforcement, so far, has landed on the side I trust. The first DSA fine (€120 million against X in December 2025) concerned no disinformation at all. It punished three transparency failures: a pay-to-play “verified” badge that misleads users, an unusable advertising repository, and X’s blocking of researchers’ access to its public data.24 The debut penalty was, in part, for sabotaging the very transparency I favor, not for refereeing opinions. For now the Ministry-of-Truth scenario is a worry about latent power, not a documented practice though X is contesting the fine, arguing the Commission writes, investigates, and judges its own rules at once.25

X is not alone: by late 2025 the Commission had opened researcher-access cases against four platforms — X, TikTok, Facebook and Instagram — and every one concerns transparency, not the content of anyone’s opinions.

Sources

  1. Sagan, C. (1979). Broca’s Brain: Reflections on the Romance of Science. Random House. ↩︎

  2. Lepenos T, Sanoudou D, Protogerou A, et al. (June 17, 2024) Louis Pasteur (1822–1895), Ignaz Semmelweis (1818–1865), Joseph Lister (1827–1912) and the Link Between Their Works Toward the Development of Antisepsis: A Narrative Review. Cureus, 16(6), e62543. doi:10.7759/cureus.62543 ↩︎

  3. Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701–721. doi:10.1037/0022-3514.79.5.701 ↩︎

  4. Small, D. A., Loewenstein, G., & Slovic, P. (2007). Sympathy and callousness: The impact of deliberative thought on donations to identifiable and statistical victims. Organizational Behavior and Human Decision Processes, 102(2), 143–153. doi:10.1016/j.obhdp.2006.01.005 ↩︎

  5. Fazio, L. K., Brashier, N. M., Payne, B. K., & Marsh, E. J. (2015). Knowledge does not protect against illusory truth. Journal of Experimental Psychology: General, 144(5), 993–1002. doi:10.1037/xge0000098 ↩︎

  6. Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151. doi:10.1126/science.aap9559. (Single-platform observational study; effect sizes are correlational.) ↩︎

  7. Mehra, M. R., Desai, S. S., Ruschitzka, F., & Patel, A. N. (2020). Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis. The Lancet (published 22 May 2020; retracted 4 June 2020). Retraction notice doi:10.1016/S0140-6736(20)31324-6. ↩︎

  8. U.S. Centers for Disease Control and Prevention (2021). Rapid Increase in Ivermectin Prescriptions and Reports of Severe Illness Associated with Use of Products Containing Ivermectin to Prevent or Treat COVID-19. CDC Health Alert Network, HAN-449, 26 August 2021. ↩︎

  9. Graff, D. M., Thompson, A. J., Berriochoa, J. P., Kuhn, M. W., Ruha, A.-M., & Lipinski, J. (2021). Chloroquine ingestion to prevent SARS-CoV-2 infection: a report of two cases. Clinical Practice and Cases in Emergency Medicine, 5(2), 161–164. doi:10.5811/cpcem.2021.3.51329 ↩︎

  10. Chang, A., Schnall, A. H., Law, R., et al. (2020). Cleaning and disinfectant chemical exposures and temporal associations with COVID-19 – National Poison Data System, United States, January 1, 2020–March 31, 2020. MMWR Morbidity and Mortality Weekly Report, 69(16), 496–498. doi:10.15585/mmwr.mm6916e1 ↩︎

  11. Johnson, S. B., Park, H. S., Gross, C. P., & Yu, J. B. (2018). Use of alternative medicine for cancer and its impact on survival. Journal of the National Cancer Institute, 110(1), 121–124. doi:10.1093/jnci/djx145 ↩︎

  12. Johnson, S. B., Park, H. S., Gross, C. P., & Yu, J. B. (2018). Complementary medicine, refusal of conventional cancer therapy, and survival among patients with curable cancers. JAMA Oncology, 4(10), 1375–1381. doi:10.1001/jamaoncol.2018.2487. ↩︎

  13. Chigwedere, P., Seage, G. R., Gruskin, S., Lee, T.-H., & Essex, M. (2008). Estimating the lost benefits of antiretroviral drug use in South Africa. JAIDS Journal of Acquired Immune Deficiency Syndromes, 49(4), 410–415. doi:10.1097/QAI.0b013e31818a6cd5. ↩︎

  14. Jia, K. M., Hanage, W. P., Lipsitch, M., et al. (2023). Estimated preventable COVID-19-associated deaths due to non-vaccination in the United States. European Journal of Epidemiology, 38, 1125–1128. doi:10.1007/s10654-023-01006-3 ↩︎

  15. Loomba, S., de Figueiredo, A., Piatek, S. J., de Graaf, K., & Larson, H. J. (2021). Measuring the impact of COVID-19 vaccine misinformation on vaccination intent in the UK and USA. Nature Human Behaviour, 5, 337–348. doi:10.1038/s41562-021-01056-1 ↩︎

  16. U.S. CDC, Measles Cases and Outbreaks (full-year 2025 total of 2,288 confirmed cases, three deaths; accessed 2026); NHS England Digital, Childhood Vaccination Coverage Statistics, England, 2023–24 (17 September 2024; MMR1 coverage at 24 months 88.9%, lowest since 2009–10); UK Health Security Agency / BMJ reporting of 2,911 confirmed measles cases in England in 2024 (highest in over two decades), with a child death in July 2025; WHO, Measles – United States of America, Disease Outbreak News, 11 March 2025. ↩︎

  17. Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A. A., Eckles, D., & Rand, D. G. (2021). Shifting attention to accuracy can reduce misinformation online. Nature, 592, 590–595. doi:10.1038/s41586-021-03344-2. The COVID-specific precursor largely failed a preregistered direct replication (the first stage was null, p = .67; a small effect emerged only when samples were pooled): Roozenbeek, J., Freeman, A. L. J., & van der Linden, S. (2021). How accurate are accuracy-nudge interventions? A preregistered direct replication of Pennycook et al. (2020). Psychological Science, 32(7), 1169–1178. doi:10.1177/09567976211024535 ↩︎

  18. Regulation (EU) 2022/2065 of 19 October 2022 on a Single Market for Digital Services (Digital Services Act), OJ L 277, 27 October 2022. In force from 16 November 2022; fully applicable from 17 February 2024 (very large platforms and search engines designated from April 2023). VLOP/VLOSE threshold: 45 million average monthly active EU users. European Commission, The Digital Services Act package, digital-strategy.ec.europa.eu. ↩︎

  19. DSA Article 27 (recommender-system transparency, all online platforms) and Article 38 (VLOPs/VLOSEs must offer at least one recommender option not based on profiling within the meaning of GDPR Art. 4(4)). Compliance is verified through independent annual audits under Article 37. ↩︎

  20. DSA Article 40 (data access for vetted researchers). European Commission, “Commission adopts delegated act on data access under the Digital Services Act,” 2 July 2025; the DSA Data Access Portal opened on 28 October 2025. ↩︎

  21. The request — by a Dutch research team seeking TikTok data to study the monetisation of political content during Romania’s 2024 presidential election as a possible Art. 34(1)(c) systemic risk — was submitted on 28 October 2025 (the day the DSA Data Access Portal opened) and rejected at the vetting stage by the Irish Digital Services Coordinator (Coimisiún na Meán) on 19 February 2026, having failed five of seven requirements. C. Goanta & A. Iamnitchi, “If at first you don’t succeed: Reflections on a rejected Art. 40 DSA data access request,” DSA Observatory, 12 March 2026. ↩︎

  22. DSA Articles 34 (systemic-risk assessment) and 35 (mitigation); Art. 34(1)(c) covers “any actual or foreseeable negative effects on civic discourse and electoral processes, and public security.” The Code of Practice on Disinformation (2018, strengthened 2022) was integrated as a Code of Conduct under Article 45, becoming a relevant compliance benchmark from 1 July 2025; adherence remains voluntary, though Recital 104 allows an unexplained refusal to participate to be weighed. European Commission, “The Code of Conduct on Disinformation,” 13 February 2025. ↩︎

  23. Husovec, M. (2024). The Digital Services Act’s red line: what the Commission can and cannot do about disinformation. Journal of Media Law, 16(1), 47–56. doi:10.1080/17577632.2024.2362483. Husovec argues the “Ministry of Truth” fear is overstated because the Commission cannot create binding rules ordering platforms to suppress lawful expression; mitigation may take the form of labelling, friction or demonetisation rather than removal. For the opposing, maximalist critique, see U.S. House Committee on the Judiciary, interim staff report on the DSA, 25 July 2025. ↩︎

  24. European Commission, “Commission fines X €120 million under the Digital Services Act,” press release, 5 December 2025 — the first non-compliance decision under the DSA, for deceptive design of the “blue checkmark,” an insufficiently transparent advertising repository, and failure to provide researchers access to public data. See also Goodwin and eucrim analyses (December 2025–January 2026). In parallel, the Commission accepted binding commitments from TikTok on its ad repository, with no fine. ↩︎

  25. X filed a challenge before the General Court of the EU in February 2026 (X v. European Commission), arguing inter alia that the Commission’s combined power to make, investigate and sanction breaches of its own rules lacks meaningful checks and balances. The case is pending. ↩︎