Homo Economicus on trial: what neuroeconomics reveals about rational choice

Eco­nom­ic the­ory has long described agents cap­able of optim­ising their choices based on com­plete inform­a­tion and coher­ent cal­cu­la­tion. Empir­ic­al obser­va­tions, both in the labor­at­ory and in the real eco­nomy, have gradu­ally qual­i­fied this view. For example, the dra­mat­ic fluc­tu­ations observed in GameStop shares in 2021 high­lighted dynam­ics of herd beha­viour and col­lect­ive over­con­fid­ence that are dif­fi­cult to recon­cile with the hypo­thes­is of inform­a­tion­al effi­ciency ¹.

Also of note, the auto­mat­ic enrol­ment of employ­ees in pen­sion schemes, which is par­tic­u­larly wide­spread in the United States fol­low­ing the Pen­sion Pro­tec­tion Act of 2006, is based on research demon­strat­ing the decis­ive influ­ence of default options on sav­ing beha­viour². Pub­lic author­it­ies also use mes­sages based on social norms or fram­ing to encour­age vac­cin­a­tion or reduce energy con­sump­tion ³.
 
These devel­op­ments are fur­ther explored in the work of Mar­ie-Claire Vil­l­ev­al, a CNRS research dir­ect­or in the GATE Lyon Saint-Étienne, where she has played a cent­ral role in the devel­op­ment of exper­i­ment­al eco­nom­ics in France. Inter­na­tion­ally recog­nised for her research in beha­vi­our­al eco­nom­ics, she stud­ies the influ­ence of social pref­er­ences, incent­ives and mor­al norms on indi­vidu­al choices. Her pub­lic­a­tions appear in the discipline’s lead­ing journ­als, includ­ing key gen­er­al­ist journ­als. By found­ing and lead­ing the GATE-Lab, an exper­i­ment­al plat­form com­bin­ing beha­vi­our­al and physiolo­gic­al meas­ure­ments, she has helped devel­op tools designed to bet­ter identi­fy the cog­nit­ive pro­cesses involved in eco­nom­ic decision-mak­ing.
 
The study of cog­nit­ive biases is no longer lim­ited to a the­or­et­ic­al chal­lenge to ration­al­ity. It addresses con­crete choices regard­ing sav­ing, reg­u­la­tion, organ­isa­tion­al gov­ernance and pub­lic policy, and raises ques­tions about how to man­age beha­viours influ­enced by mech­an­isms that are some­times implicit.

FALSE

The design of an exper­i­ment­al pro­tocol begins with a the­or­et­ic­al mod­el. First, a stand­ard frame­work of ration­al and self-inter­ested indi­vidu­als is employed, fol­lowed by a beha­vi­our­al mod­el incor­por­at­ing cog­nit­ive biases and social pref­er­ences. This the­or­et­ic­al frame­work enables the for­mu­la­tion of hypo­theses, the defin­i­tion of rel­ev­ant vari­ables and, through appro­pri­ate exper­i­ment­al treat­ments, the isol­a­tion of cer­tain mech­an­isms. To meas­ure biases such as loss aver­sion, present bias or risk aver­sion, indi­vidu­al decision-mak­ing games are favoured. These biases are gen­er­ally insens­it­ive to the inten­tions or gains of oth­ers. Con­versely, the study of social pref­er­ences requires inter­ac­tions between play­ers to identi­fy altru­ism, aver­sion to inequal­ity or sens­it­iv­ity to norms.

The meth­od­o­lo­gic­al chal­lenge lies in manip­u­lat­ing these dimen­sions ortho­gon­ally to identi­fy their respect­ive con­tri­bu­tions. Loss aver­sion refers to sens­it­iv­ity to one’s own losses, whilst social pref­er­ences cap­ture sens­it­iv­ity to oth­ers’ gains and losses. It is, how­ever, pos­sible that cog­nit­ive biases and social pref­er­ences are cor­rel­ated. Loss aver­sion may rein­force con­cerns about fair­ness, and mor­al pref­er­ences may define the ref­er­ence point from which a situ­ation is per­ceived as a loss or a gain. A con­trolled pro­tocol stand­ard­ises the decision-mak­ing situ­ation to lim­it com­pet­ing factors. When sev­er­al dimen­sions are likely to inter­act, addi­tion­al data is col­lec­ted at the end of the ses­sion, such as on atti­tudes towards risk, patience or mor­al pref­er­ences, which may influ­ence the observed behaviours.

UNCLEAR

Com­bin­ing beha­vi­our­al meas­ure­ments with physiolo­gic­al or brain data helps to shed light on the mech­an­isms under­ly­ing choices. Eye-track­ing provides inform­a­tion on the alloc­a­tion of atten­tion and helps identi­fy inform­a­tion avoid­ance or the sali­ence effect. In a shar­ing game, observing wheth­er an indi­vidu­al sys­tem­at­ic­ally ignores oth­ers’ gains helps identi­fy selfish motiv­a­tion. It can also reveal the avoid­ance of mor­ally bind­ing information.

Elec­tro­physiolo­gic­al meas­ures, such as skin con­duct­ance or heart rate, shed light on the role of emo­tions, their intens­ity and their valence. They help dis­tin­guish between dif­fer­ent motiv­a­tions for cooper­a­tion, such as the pur­suit of effi­ciency, reci­pro­city, group mem­ber­ship or shame asso­ci­ated with an insuf­fi­cient con­tri­bu­tion. Neur­al data from EEG or fMRI can identi­fy the activ­a­tion of areas involved in reward, dis­gust or value cal­cu­la­tion, not­ably the stri­atum. They can link beha­vi­our­al vari­ables to bio­lo­gic­al traces and detect uncon­scious pro­cesses, such as risk anti­cip­a­tion or pre­dic­tion errors in learning.

These approaches do, how­ever, have lim­it­a­tions. They are costly, often restrict the obser­va­tion of mul­tiple inter­ac­tions and com­plic­ate the col­lec­tion of large samples, where­as much research focuses on small effects requir­ing high stat­ist­ic­al power. They gen­er­ally do not allow for caus­al iden­ti­fic­a­tion, unlike beha­vi­our­al protocols.

TRUE

Mon­et­ary incent­ives play a cent­ral role in exper­i­ment­al eco­nom­ics, as they allow val­ues to be induced in a con­trolled man­ner. Their effect on biases is, how­ever, mixed. They can reduce cer­tain biases by increas­ing atten­tion and engage­ment, par­tic­u­larly inat­ten­tion and anchor­ing biases. High fin­an­cial stakes tend to increase cau­tion and thus risk aver­sion. Con­versely, present bias, loss aver­sion or truth pref­er­ence show little response to changes in incent­ives, sug­gest­ing that they stem from struc­tur­al char­ac­ter­ist­ics of pref­er­ences or heur­ist­ics that are dif­fi­cult to neutralise.

Incent­ives can also pro­duce effects con­trary to their object­ive by redu­cing intrins­ic motiv­a­tion, par­tic­u­larly in the mor­al or social sphere. A tax incent­ive for pur­chas­ing an elec­tric car may blur the proso­cial sig­ni­fic­ance of the choice and dis­cour­age those whose motiv­a­tion is intrins­ic. Sim­il­arly, a fine for being late to the nurs­ery may trans­form a rela­tion­al norm into a com­mer­cial trans­ac­tion and increase lateness. 

At very high levels, incent­ives can rein­force over­con­fid­ence, the illu­sion of con­trol and excess­ive com­pet­it­ive­ness, or impair per­form­ance due to the stress of high stakes. These con­trast­ing effects call for cau­tion in the design of pub­lic or private incent­ive mechanisms.

FALSE

From the stand­ard per­spect­ive, pun­ish­ing oth­ers at per­son­al cost is irra­tion­al. How­ever, exper­i­ments show that indi­vidu­als are will­ing to pun­ish those who devi­ate from group norms, even sym­bol­ic­ally. This beha­viour reflects social pref­er­ences, par­tic­u­larly an aver­sion to inequal­ity and injustice, rather than a cog­nit­ive bias. Loss aver­sion may, how­ever, explain a stronger reac­tion to pun­ish­ment than to a reward. 

The per­cep­tion of sanc­tions depends on pro­ced­ur­al justice and trans­par­ency. A mech­an­ism adop­ted by vote is more widely accep­ted and may even reduce the need to apply sanc­tions by mak­ing the norms expli­cit. Con­versely, sanc­tions per­ceived as arbit­rary can trig­ger revenge and spir­als of counter-pun­ish­ment, espe­cially in unequal con­texts. “Anti­so­cial” sanc­tions dir­ec­ted against cooper­at­ors also exist and under­mine cooper­a­tion; they appear to be more com­mon in cer­tain cul­tures, par­tic­u­larly in former com­mun­ist bloc countries.

Finally, inform­al sanc­tions aim to uphold group norms. When norms that have become obsol­ete per­sist due to “plur­al­ist­ic ignor­ance,” pref­er­ences for change remain unex­pressed, gen­er­at­ing frus­tra­tion and with­draw­al. An expli­cit col­lect­ive expres­sion of pref­er­ences, for example through a ref­er­en­dum, can cre­ate a shared under­stand­ing of the desire for change and revive cooperation.

UNCLEAR

Extern­al valid­ity requires the rep­lic­a­tion of exper­i­ments in var­ied con­texts, with dif­fer­ent pop­u­la­tions, stakes, and cul­tures, fol­low­ing a meta-sci­entif­ic approach. Even in the absence of a per­fect cor­rel­a­tion between the labor­at­ory and every­day life, many biases observed in con­trolled envir­on­ments are found in real-world situ­ations. Loss aver­sion and over­con­fid­ence appear in invest­ment decisions; the over­em­phas­is on small prob­ab­il­it­ies explains lot­tery play or over­insur­ance; the pref­er­ence for the present mani­fests itself in pro­cras­tin­a­tion regard­ing work, stud­ies, or health.

In every­day life, indi­vidu­als have more time to learn or avoid cer­tain situ­ations, provided they are aware of their biases. This raises a norm­at­ive ques­tion: to what extent can pub­lic author­it­ies inter­vene to cor­rect lim­ited and con­tex­tu­al ration­al­ity? Nudges can be per­ceived as bene­vol­ent assist­ance or as an intru­sion. Their accept­ab­il­ity var­ies across cul­tures and pref­er­ences regard­ing the legit­im­acy of pub­lic action.

Aicha Fall