Are cognitive biases compatible with the scientific method?

Cog­ni­tive bias­es are well doc­u­ment­ed in cog­ni­tive sci­ence research¹. These sys­tem­at­ic – and there­fore pre­dictable – errors are not only a sign of our lim­it­ed ratio­nal­i­ty, but they also explain the way our judg­ments and deci­sions work. As such, cog­ni­tive bias­es are a group of inher­ent process­es con­trol­ling the mind so that it can: man­age large flows of infor­ma­tion, com­pen­sate for the lim­its of mem­o­ry, pre­serve cog­ni­tive econ­o­my, make quick deci­sions, access mean­ing­ful expla­na­tions, pro­tect integri­ty of the self and reas­sure our­selves in our decisions.

Study­ing cog­ni­tive bias­es in a sci­en­tif­ic way con­sists of build­ing up a ratio­nal body of knowl­edge to bet­ter under­stand our irra­tional­i­ty. To do this, the sci­en­tif­ic method relies on the one hand on the descrip­tion of objec­ti­fi­able facts, which we iden­ti­fy using quan­ti­tive meth­ods. And on the oth­er, by using invari­ant explana­to­ry mod­els (a.k.a. the­o­ries) – which must cor­re­spond with known facts and be sub­se­quent­ly used to pre­dict, test, and com­pare – through which we seek to under­stand the causal­i­ty of phe­nom­e­na made pos­si­ble by an exper­i­men­tal method².

Since the truth is not always easy to find, sci­ence is full of con­tro­ver­sies. This is why the sci­en­tif­ic method is based on the fun­da­men­tal prin­ci­ple of “dis­pute”, i.e. debate of the results obtained, between peers, with demon­strat­ed pub­licly-avail­able proof. It is there­fore col­lec­tive, sub­ject to crit­i­cism and repli­ca­tion, with nuance, over a long peri­od of time and inde­pen­dent from polit­i­cal influ­ence so that we may con­verge towards truth.

How­ev­er, it should be said that the ingre­di­ents of the sci­en­tif­ic method and cog­ni­tive bias­es are some­times (or even, often) antag­o­nis­tic. With­out claim­ing to be exhaus­tive, let’s iden­ti­fy some sig­nif­i­cant stum­bling blocks that may help us to bet­ter under­stand cer­tain con­tem­po­rary issues around mis­trust in science.

Imag­ine that I have a rule in mind that I sug­gest you guess. I inform you that the sequence of num­bers “2, 4 and 6” respects this rule. To guess it, you can pro­pose oth­er sequences of three num­bers, and I will tell you if they con­form to my rule or not. When we car­ry out this exper­i­ment³, the par­tic­i­pants will log­i­cal­ly make a hypoth­e­sis about the rule (for exam­ple “sequence of num­bers increas­ing by two each time”), and test it pos­i­tive­ly, with a large major­i­ty of jus­ti­fi­ca­tion series such as “16, 18, 20” and then “23, 25, 27”.

The pur­pose of these con­fir­ma­to­ry state­ments is not to test IF the hypoth­e­sis is true, but THAT the hypoth­e­sis is true. Only the series that will inval­i­date the hypoth­e­sis for­mu­lat­ed by the par­tic­i­pants (e.g. here “3, 6, 9”) will make it pos­si­ble to ver­i­fy IF it is true. This “hypoth­e­sis con­fir­ma­tion bias” explains why we spon­ta­neous­ly and care­ful­ly avoid look­ing for argu­ments that go against our beliefs: the aver­sion to los­ing our cer­tain­ties out­weighs the pos­si­bil­i­ty of gain­ing new knowl­edge. As some­one once said, “Insan­i­ty is doing the same thing over and over again and expect­ing a dif­fer­ent result”.

We tend to over­es­ti­mate the prob­a­bil­i­ty of an event when we know that it has tak­en place.

The sci­en­tif­ic method, on the oth­er hand, is counter-intu­itive, and teach­es us to beware of this bias thanks to the dou­ble-blind tech­nique designed to lim­it self-per­sua­sion, and to an “infir­ma­to­ry” pos­ture: test­ing hypothe­ses by mul­ti­ply­ing exper­i­ments like­ly to refute them. Hence, a the­o­ry “resists” the facts until proven oth­er­wise. Nev­er­the­less, the process of research is not entire­ly free from con­fir­ma­tion bias because pos­i­tive results are con­sid­er­ably well-val­ued by pub­li­ca­tions, espe­cial­ly in the so-called “social sci­ences”. More­over, repro­ducibil­i­ty stud­ies are not always pop­u­lar, espe­cial­ly when they reveal how many research results in the human­i­ties and social sci­ences can­not be reproduced[/pi_note]Larivée, S., Sénéchal, C., St-Onge, Z. & Sauvé, M.-R. (2019). « Le biais de con­fir­ma­tion en recherche ». Revue de psy­choé­d­u­ca­tion, 48(1), 245–263[/pi_note].

The pow­er of hypoth­e­sis con­fir­ma­tion bias lies in the fact that it does not only con­cern the present but also… the past! Indeed, we tend to over­es­ti­mate the prob­a­bil­i­ty of an event when we know that it has tak­en place: after the fact, we often behave as if the future were obvi­ous to pre­dict (“that was bound to hap­pen”), and as if uncer­tain­ty or the unknown did not inter­vene in the events. This “ret­ro­spec­tive” con­fir­ma­tion bias⁴ is all the more salient in trag­ic sit­u­a­tions and may explain crit­i­cism of sci­en­tists’ or politi­cians’ inten­tions once the human toll of a pan­dem­ic, a ter­ror­ist attack or an eco­nom­ic cri­sis is known.

The ret­ro­spec­tive bias relies on the extra­or­di­nary capac­i­ty of the human mind for ratio­nal­i­sa­tion, i.e. the jus­ti­fi­ca­tion of events after the fact. We can nev­er resist telling our­selves a good sto­ry, even if it means dis­tort­ing real­i­ty⁵. As a result, the fran­tic search for caus­es is pre­ferred to sim­ple cor­re­la­tions, pseu­do-cer­tain­ties to prob­a­bil­i­ties, the denial of chance to the con­sid­er­a­tion of haz­ards, dichoto­mous think­ing to nuance, the over­es­ti­ma­tion of low prob­a­bil­i­ties to the neu­tral obser­va­tion of facts: pre­cise­ly the oppo­site of what the sci­en­tif­ic method teach­es us.

Can the sci­en­tif­ic method be applied to the study of humans by humans? In a vast series of research stud­ies in exper­i­men­tal social psy­chol­o­gy, Jean-Pierre Deconchy and his team explored a fas­ci­nat­ing sub­ject: the way human­i­ty thinks about human­i­ty, and the way human­i­ty thinks about the study of human­i­ty. With the help of inge­nious exper­i­men­tal set-ups col­lect­ed in the year 2000 (pub­lished in Les ani­maux sur­naturés⁶), researchers showed how, in the absence advanced sci­en­tif­ic cul­ture, some of our cog­ni­tive fil­ters con­vince our­selves that our thoughts and behav­iours are not based on nat­ur­al deter­mi­nants. And that, con­se­quent­ly, by virtue of these cog­ni­tive fil­ters, sci­ence would be unfit to under­stand and explain deep human “nature”.

Thus, humans con­struct a def­i­n­i­tion of human­i­ty, which sep­a­rates them­selves from the idea that they are crea­tures of nature, deter­mined by the same laws as oth­er liv­ing beings. And that behind this bio­log­i­cal form hides anoth­er “thing”, a “super-nature”, and thus a defi­ance of the very idea that sci­ence has a word to say on what human­i­ty is.

In this research, we find the idea of lim­it­ed ratio­nal­i­ty, in the sense that the knowl­edge of human­i­ty would be some­thing oth­er than ratio­nal­i­ty. It is also incred­i­ble to see that, at the same time as we progress in cog­ni­tive and neu­ro­sciences, we are also wit­ness­ing sev­er­al pseu­do-human sci­ences flour­ish­es, adding a lit­tle extra soul to the “super-nature” stud­ied by Deconchy. These include a revival of shaman­ism, ener­getic ‘med­i­cine’ and per­son­al devel­op­ment tech­niques. They adopt sci­en­tif­ic vocab­u­lary that has an author­i­ta­tive (anoth­er cog­ni­tive bias) effect – some­thing that we have recent­ly seen in fan­ci­ful extrap­o­la­tions claim­ing terms from quan­tum physics to jus­ti­fy alter­na­tive med­i­cines or oth­er mys­te­ri­ous phe­nom­e­na⁷.

Our brain draws quick and cheap con­clu­sions to do us a favour. Most of the time, they are suf­fi­cient and rough­ly rel­e­vant to our imme­di­ate needs. But some­times, they do us a dis­ser­vice and lead us down a path that dis­cred­its the very idea of free will. To fight against one­self, against the nat­ur­al slope of cog­ni­tive bias­es that weak­en our dis­cern­ment, requires min­i­mal train­ing in what sci­en­tif­ic method is – not only for those who are des­tined to a sci­en­tif­ic pro­fes­sion. It also requires an under­stand­ing of the short­cuts our brain uses to make our lives eas­i­er, and some­times to lull us into an illu­sion of understanding.

Char­i­ties such as “La main à la pâte” (in France) and, more glob­al­ly, the projects ded­i­cat­ed to sci­en­tif­ic out­reach, in con­nec­tion with uni­ver­si­ties and research organ­i­sa­tions, are feed­ing a real soci­etal need to rein­force psy­cho-social skills, not only of school­child­ren but also of cit­i­zens. This is the price to pay so that sci­ence is not per­ceived as a belief like any oth­er, so that doubt­ful or mis­lead­ing opin­ions do not take prece­dence over the truth, and thus so that our democ­ra­cies main­tain their eman­ci­pa­to­ry skills.