In 1983, American sociologist W. Phillips Davison gave a name to a mechanism any careful observer already sensed, the near-universal conviction that deceptive messages affect others more than oneself. Four decades later, research on political disinformation confirms the same pattern, and a recent survey on the 2026 French municipal elections documents it with unusual precision, seventy-two out of a hundred citizens believe they can tell false from true online, while only twenty-three out of a hundred grant that same ability to their fellow citizens. Blaise Pascal reached a similar conclusion three centuries earlier, when he wrote that intelligence alone does not suffice to recognise truth unless one loves it enough to be corrected by it. The bias also carries a concrete political consequence, it inclines people to prefer that others be educated rather than regulating what anyone, themselves included, may read or share. AcidReport examines this mechanism, not as a laboratory curiosity but as the political condition of an age that mistakes repetition for evidence.
The bias Davison named in 1983
Davison called his finding the third-person effect, the tendency to assume that a persuasive or deceptive message influences others more than oneself. The sociologist first documented it in the field of wartime propaganda, but the pattern proved far broader than his original subject.
Four decades of subsequent research confirm it in contexts Davison could not have foreseen, among them digital disinformation. A national survey conducted in the United States in March 2017 by researchers S. Mo Jang and Joon K. Kim, drawing on a sample of 1,299 people, found the same bias applied to fake news. Partisan identity and a sense of personal political efficacy emerged as the main factors reinforcing the perception that other groups were more vulnerable to deception than one’s own. The mechanism depends on neither country, language, nor election type. It depends only on the existence of a message the subject deems false and a third party to whom the naivety of believing it can be attributed.
Jang and Kim also found a subtler political consequence than a simple tolerance for censorship, and in some ways a more uncomfortable one. The more strongly a person perceived the gap between their own judgement and that of others, the more they supported media literacy programmes aimed at those others, and the less they supported direct regulation of platforms or content. No one asks to be educated themselves against the lie. Almost everyone prefers that others be educated, rather than restricting what they themselves might want to read or share.
Formulated in 1983 by sociologist W. Phillips Davison, the third-person effect has been confirmed in dozens of subsequent studies, including those focused specifically on fake news and political disinformation. (Davison, The Third-Person Effect in Communication, Public Opinion Quarterly, 1983)
Who benefits from the bias
The bias benefits no one in particular while it remains private, each person convinced of their own immunity to the lie they see in others. But it becomes a political resource the moment someone learns to exploit it from outside.
A well-designed disinformation campaign does not need to convince the average citizen that the lie is true. It only needs that citizen to believe that only others, those of the other party, those of the other country, the less educated, are likely to believe it. The false message then circulates protected by everyone else’s certainty of being immune to it, with no one feeling any need to examine what they have just shared.
Davison himself opened his 1983 article with the episode that led him to the concept. On Iwo Jima, during the Second World War, a service unit made up of black soldiers under white officers received Japanese leaflets urging them not to risk their lives for the white man and to surrender or desert at the first opportunity. Nothing suggested the message had affected the soldiers it was addressed to. It affected, instead, the white officers, who withdrew the unit the following day. The Japanese propagandist did not need to convince the leaflet’s direct recipient. It was enough that a third party believed that recipient could be convinced.
This is, in good part, why media literacy campaigns, funded almost always by institutions aimed at the general public, yield modest results. They teach people to recognise the hoax reasonably well, but do not touch the psychological mechanism that makes the reader feel exempt from applying what they learned to their own consumption of information. One can know, in the abstract, how to spot a fake news story, and still fail to apply that knowledge to the message one has just forwarded to a family group chat.
A recent, well-documented case
Almost any current survey on disinformation can illustrate the bias, but few do so with the precision the Fondation Jean-Jaurès brought to the 2026 French municipal elections. Seventy-two out of a hundred respondents claim they can recognise false information as soon as they see it. Asked about that same ability in their fellow citizens, the figure collapses to twenty-three.
The gap between those two numbers does not measure other people’s gullibility. It measures one’s own blindness, and that is where the real interest of the data lies. This asymmetry is no statistical accident, it is the very mechanism Davison described in 1983, updated with measurement tools he did not have. If everyone believes themselves immune, no one watches their own consumption of information, and vigilance falls, by default, to others, who are just as convinced of their own immunity.
The same survey also details where the lie circulates and who originates it, according to respondents themselves. 51% of those who reported having been exposed to false information about local life received it in private conversation, against 41% who received it on social media and only 22% who spotted it in the local press. Asked about its origin, 44% pointed to other ordinary citizens, 39% to local candidates or political figures, and 26% to activists or militants. 47% of respondents also declare themselves worried about possible foreign interference in the municipal election process, a fear the scale of the phenomenon does not by itself suffice to justify.
72% of respondents say they can distinguish true information from false, but only 23% grant that same ability to the rest of the population. (Fondation Jean-Jaurès, survey on the 2026 French municipal elections)
The same logic of underestimation repeats at another level. 39% of respondents consider municipal elections less exposed to disinformation than national ones, against only 9% who judge them more exposed. Proximity, paradoxically, reassures. People watch the distant discourse carefully, the presidential campaign, the international controversy, and lower their guard against rumours about the mayor or the local public works project, precisely the terrain where other people’s word carries the most weight because it circulates among people who know each other.
What Pascal already sensed
Blaise Pascal had neither empirical sociology nor opinion polls at his disposal, but he reached a conclusion Davison would take three centuries to formulate with method. He wrote, in a fragment of his Pensées composed around 1660, that truth is so obscured in this age and falsehood so established that, unless one loves the truth, one cannot know it.
The two are not saying exactly the same thing, and it is worth stating plainly. Pascal describes a moral disposition, the willingness to be corrected, within an unfinished theological project. Davison describes a cognitive bias, measurable and replicated in the laboratory, unrelated to any question of faith. But both point to the same blind spot, the resistance to examining one’s own judgement with the same rigour applied to that of others.
The same Pascal who wrote about the love of truth had spent much of his adult life building calculating machines, studying atmospheric pressure, and laying the foundations of probability theory. He was no stranger to rigorous reasoning, quite the opposite, he was one of the most gifted practitioners of his century. That such a man concluded intelligence is not enough is not the confession of someone incapable of reasoning well. It is the observation of someone who reasoned better than almost anyone and who, precisely for that reason, could see the exact limit of what reasoning achieves on its own.
Pascal was also writing as part of an unfinished project, an apology for Christianity that death, in 1662, kept him from completing and that his heirs published eight years later under the title by which it is known today. The fragment on obscured truth speaks of no technology, no survey, no algorithm. It speaks of a disposition of the soul, the willingness to be corrected, which Pascal considered rarer than intelligence itself and, unlike intelligence, not distributed by chance but chosen.
The sentence is uncomfortable because it inverts the expected order. It does not say more information is needed, better verification tools, a more vigilant press, though all of that helps. It says wanting comes before knowing.
Intelligence spots the other person’s error with almost athletic ease. It finds the fallacy in a political opponent’s argument, the manipulation in the headline of a disliked outlet, the lie in the mouth of someone already suspected. That same intelligence grows noticeably slower when the error lies in one’s own conviction. It is not that capacity is lacking. Disposition is lacking, and disposition is not measured in IQ, it is measured in the willingness to lose an argument against oneself.
The cost of one’s own exemption
To love the truth, in the sense Pascal gave the word, is not to possess it. It is to accept that it may strip us of something, a comfortable prejudice, a certainty that propped up half an identity, the flattering image each of us patiently builds of ourselves over the years. The lie demands no such sacrifice. It offers, in exchange for submission, the effortless coherence of a world where one’s own convictions are never wrong.
That is its decisive advantage, and it explains why it thrives even among people who consider themselves reasonable, informed, immune. The third-person effect distinguishes neither nationality nor level of education, every new survey that dares ask confirms it, not whether the respondent believes the lie, but whether they believe others believe it. No one thinks they belong to the mistaken majority. Everyone, without exception, thinks they are the exception.
No institutional reform resolves by decree what is, above all, an intimate economy of self-esteem. A platform can be regulated, an outlet sanctioned, a hoax debunked with impeccable data, and the mechanism will remain intact as long as everyone keeps the comfortable certainty that the problem starts on the neighbour’s screen.
Journalism itself does not escape the trap, and it would be convenient to pretend otherwise. No newsroom believes itself manipulable, every one of them suspects the newsroom across the street. The same vigilance demanded of the distracted reader rarely applies, with equal rigour, to its own sources. Presenting this text as the exception that escapes the bias it describes would, in fact, be the most elegant way of falling straight into it.
That is where the only available resistance begins, not in blanket suspicion of everything one reads or hears, which is merely another form of intellectual comfort, but in the willingness to doubt one’s own certainty first, before doubting that of others. Pascal did not ask for intelligence. He asked for something harder to fake…
G.S.
Sources
- La vérité est si obscurcie en ce temps (Pensées, Brunschvicg 341, Lafuma 736, Sellier 617)
- Municipales 2026, la menace discrète de la désinformation (Fondation Jean-Jaurès)
- The Third-Person Effect in Communication, W. Phillips Davison, Public Opinion Quarterly, 1983 — archive unavailable at time of publication, source inaccessible
- Third person effects of fake news, fake news regulation and media literacy interventions, S. Mo Jang and Joon K. Kim, Computers in Human Behavior, vol. 80, 2018 — archive unavailable at time of publication, source inaccessible
