Intuition and Unconscious Learning

Part of the se­quence: Ra­tion­al­ity and Philosophy

We have already ex­am­ined two sources of our in­tu­itions: the at­tribute sub­sti­tu­tion heuris­tics and our evolved psy­chol­ogy. To­day we look at a third source of our in­tu­itions: un­con­scious learn­ing.

Un­con­scious learning

The ‘learn­ing per­spec­tive’ on in­tu­ition is com­pat­i­ble with the heuris­tics and bi­ases liter­a­ture and with evolu­tion­ary psy­chol­ogy, but adds a deeper un­der­stand­ing of what is go­ing on ‘un­der the hood.’ The learn­ing per­spec­tive says that many in­tu­itions rely on rep­re­sen­ta­tions that re­flect the en­tirety of ex­pe­riences stored in long-term mem­ory. Such in­tu­itions merely re­pro­duce statis­ti­cal reg­u­lar­i­ties in long-term mem­ory.1

An ex­am­ple will help ex­plain:

As­sume you run into a man at the 20th an­niver­sary party of your high school class grad­u­a­tion. You im­me­di­ately sense a feel­ing of dis­like. To avoid get­ting into a con­ver­sa­tion, you sig­nal and shout some words to a cou­ple of old friends sit­ting at a dis­tant table. While you are walk­ing to­ward them, you try to re­mem­ber the man’s name, which pops into your mind af­ter some time; and sud­denly, you also re­mem­ber that it was he who always did nasty things to you such as tak­ing your se­cret let­ters and show­ing them to the rest of the class. You ap­plaud the product of your in­tu­ition (the im­me­di­ate feel­ing) that has helped you to make the right de­ci­sion (avoid­ing in­ter­ac­tion). Re­call of prior ex­pe­riences was not nec­es­sary to make this de­ci­sion. The de­ci­sion was solely based on a feel­ing, which re­flected prior knowl­edge with­out aware­ness.2

Learn­ing per­spec­tive the­o­rists would sug­gest that your feel­ing of dis­like—your in­tu­ition that you shouldn’t talk to the man—came from some­thing like an (un­con­scious) reg­u­lar­i­ties anal­y­sis of your ex­pe­riences with that man that were stored in long-term mem­ory, and those ex­pe­riences turned out to be mostly nega­tive. As such, your in­tu­ition can make use of rapid par­allel pro­cess­ing to draw on the whole sum of ex­pe­riences in long-term mem­ory, rather than us­ing a slower, se­quen­tial-pro­cess­ing judg­ment al­gorithm.

It is difficult to track the source of any par­tic­u­lar in­tu­ition (though we can try3), but there is ev­i­dence to sug­gest that un­con­scious learn­ing is a com­mon source of our in­tu­itions.

Stock tickers

In a se­ries of ex­per­i­ments,4 re­searchers asked sub­jects to watch a se­ries of ad­ver­tise­ments. They warned sub­jects that a (fic­tional) stock ticker at the bot­tom of the screen would be added as a dis­trac­tor (screen­shot), and that they would be quizzed on the ad­ver­tise­ments later. After be­ing quizzed on the ad­ver­tise­ments, sub­jects were sur­prised by a quiz on their at­ti­tudes to­ward the fic­tional stocks. Post-ex­per­i­ment in­ter­views con­firmed that sub­jects had not in­tended to form at­ti­tudes to­ward the stocks.

Sub­jects watched 20 to 40 ad­ver­tise­ments while the ‘dis­trac­tor’ stock ticker dis­played 70 to 140 re­turn val­ues for 4 to 8 shares. As the in­de­pen­dent vari­able, re­searchers varied the re­turn val­ues (and thus their sum, av­er­age, fre­quency, and peaks).

When given the sur­prise quiz on their at­ti­tudes to­ward the fic­tional stocks, re­searchers found a perfect rank cor­re­la­tion be­tween the sub­jects’ mean eval­u­a­tion of the shares and the sums of their re­turns. This was the case even though sub­jects had no con­crete mem­o­ries of the share re­turns, and could not re­mem­ber the sum or av­er­age val­ues. Sub­jects re­ported they had re­lied on their ‘gut re­ac­tion’ or ‘in­tu­itive feel­ing.’

Here, it does not seem that sub­jects were able to ar­rive at such ac­cu­rate in­tu­itions by way of a spe­cific evolved in­tu­ition or an at­tribute sub­sti­tu­tion heuris­tic. In­stead, they seem to have drawn upon their un­con­scious learn­ing sys­tem with­out know­ing that they were do­ing so.

Base rate neglect

Con­sider this prob­lem:

If a test to de­tect a dis­ease whose prevalence is 1/​1000 has a false pos­i­tive rate of 5%, what is the chance that a per­son found to have a pos­i­tive re­sult ac­tu­ally has the dis­ease, as­sum­ing you know noth­ing about the per­son’s symp­toms or signs?5

Among 60 Har­vard med­i­cal stu­dents and staff, al­most half judged that the per­son has the dis­ease with .95 prob­a­bil­ity, while only 18% got the cor­rect an­swer: .02. This is an ex­am­ple of base rate ne­glect. Sub­jects based their judg­ment mostly on the ev­i­dence from the test, and ig­nored the strong ev­i­dence from the base rate (1/​1000).

Base rate sen­si­tivity im­proves when such prob­lems are framed in terms of fre­quen­cies rather than prob­a­bil­ities, but even then base rate ne­glect oc­curs in about half of sub­jects.6

Sub­jects fur­ther im­prove their statis­ti­cal judg­ments when they are al­lowed learn the dis­tri­bu­tion of a vari­able by their own sam­pling, and be­come even more sen­si­tive to base rates.7

In a re­lated study,8 re­searchers had sub­jects perform sev­eral be­hav­iors many times, and then asked them to es­ti­mate be­hav­ior fre­quency. Half of the sub­jects were asked to make spon­ta­neous judg­ments, and half were asked to de­liber­ate care­fully about their judg­ments. In the de­liber­a­tion con­di­tion, judg­ments were bi­ased by the availa­bil­ity heuris­tic. Judg­ments from the spon­ta­neous judg­ment con­di­tion were more ac­cu­rate, and seemed to re­flect un­con­scious re­call of the to­tal­ity of be­hav­iors just performed, stored by un­con­scious learn­ing.

Conclusion

Th­ese and many oth­ers stud­ies9 sug­gest that some­times our feel­ings and in­tu­itive judg­ments arise from un­con­scious par­allel pro­cess­ing of all (or many) of the ex­pe­riences rele­vant to a given judg­ment stored in our long-term mem­ory.

Later we’ll ex­am­ine how this un­der­stand­ing of in­tu­ition (along with the per­spec­tives from at­tribute sub­sti­tu­tion heuris­tics and evolu­tion­ary psy­chol­ogy) gives us some clues about how much trust we should put in our in­tu­itions un­der par­tic­u­lar con­di­tions, and how we can train our in­tu­itions to be more ac­cu­rate.10

Next post: When In­tu­itions Are Useful

Pre­vi­ous post: Your Evolved Intuitions

Notes

1 Betsch et al. (2004); Betsch & Haber­stroh (2005); Betsch (2007); Klein (1999); Hog­a­rth (2001, 2007); Ep­stein (2007). For an overview of the neu­ro­science of un­con­scious learn­ing, see Volz & Cra­mon (2007). For an overview of the re­la­tion be­tween emo­tion and in­tu­ition, see Zee­len­berg et al. (2007).

2 Betsch (2007), p. 6.

3 Hamm (2007).

4 Betsch et al. (2001, 2003, 2007).

5 Tver­sky & Kah­ne­man (1982), p. 154.

6 Gigeren­zer & Hoffrage (1995).

7 Betsch et al. (1998); Fiedler et al. (2000).

8 Haber­stroh et al. (2006).

9 Pless­ner et al. (2007); Raab & John­son (2007); Glöck­ner (2007). Also see re­search on the ‘sam­ple size effect’: Kauf­mann & Betsch (2009).

10 Hog­a­rth (2001, 2007); Erev et al. (2007).

References

Betsch (2007). The na­ture of in­tu­ition and its ne­glect in re­search on judg­ment and de­ci­sion mak­ing. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 3-22). Psy­chol­ogy Press.

Betsch, Pless­ner, Sch­wieren, & Gütig (2001). I like it but I don’t know why: A value-ac­count ap­proach to im­plicit at­ti­tude for­ma­tion. Per­son­al­ity and So­cial Psy­chol­ogy Bul­letin, 27: 242–253.

Betsch, Hoff­mann, Hoffrage, & Pless­ner (2003). In­tu­ition be­yond recog­ni­tion: When less fa­mil­iar events are liked more. Ex­per­i­men­tal Psy­chol­ogy, 50: 49–54.

Betsch, Pless­ner, & Schal­lies (2004). The value-ac­count model of at­ti­tude for­ma­tion. In Had­dock & Miao (eds.), Con­tem­po­rary per­spec­tives on the psy­chol­ogy of at­ti­tudes (pp. 251-274). Psy­chol­ogy Press.

Betsch & Haber­stroh, eds. (2005). The rou­tines of de­ci­sion mak­ing. Psy­chol­ogy Press.

Betsch, Kauf­mann, Lin­dow, Pless­ner, & Hoff­mann (2006). Differ­ent prin­ci­ples of in­for­ma­tion in­te­gra­tion in im­plicit and ex­plicit at­ti­tude for­ma­tion. Euro­pean Jour­nal of So­cial Psy­chol­ogy, 36: 887–905.

Betsch, Biel, Ed­delbüttel, & Mock (1998). Nat­u­ral sam­pling and base-rate ne­glect. Euro­pean Jour­nal of So­cial Psy­chol­ogy, 28: 269–273.

Ep­stein (2007). In­tu­ition from the per­spec­tive of cog­ni­tive-ex­pe­ri­en­tial self-the­ory. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 23-37). Psy­chol­ogy Press.

Erav, Shi­monow­itch, Schurr, & Her­twig (2007). Base rates: How to make the in­tu­itive mind ap­pre­ci­ate or ne­glect them. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 135-148). Psy­chol­ogy Press.

Fiedler, Brinkmann, Betsch, & Wild (2000). A sam­pling ap­proach to bi­ases in con­di­tional prob­a­bil­ity judg­ments: Beyond base-rate ne­glect and statis­ti­cal for­mat. Jour­nal of Ex­per­i­men­tal Psy­chol­ogy, Gen­eral, 129: 399–418.

Gigeren­zer & Hoffrage (1995). How to im­prove Bayesian rea­son­ing with­out in­struc­tion: Fre­quency for­mats. Psy­cholog­i­cal Re­view, 102: 684–704.

Glöck­ner (2007). Does in­tu­ition beat fast and fru­gal heuris­tics? A sys­tem­atic em­piri­cal anal­y­sis. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 309-326). Psy­chol­ogy Press.

Haber­stroh, Betsch, & Aarts (2006). When guess­ing is bet­ter than think­ing: Mul­ti­ple bases for fre­quency judg­ments. Un­pub­lished manuscript.

Hamm (2007). Cue by hy­poth­e­sis in­ter­ac­tions in de­scrip­tive mod­el­ing of un­con­scious use of mul­ti­ple in­tu­itive judg­ment strate­gies. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 55-70). Psy­chol­ogy Press.

Hog­a­rth (2001). Ed­u­cat­ing In­tu­ition. Univer­sity of Chicago Press.

Hog­a­rth (2007). On the learn­ing of in­tu­ition. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 91-106). Psy­chol­ogy Press.

Klein (1999). Sources of power. How peo­ple make de­ci­sions. MIT press.

Kauf­mann & Betsch (2009). Ori­gins of the sam­ple-size effect in ex­plicit eval­u­a­tive judg­ments. Ex­per­i­men­tal Psy­chol­ogy, 56: 344-353.

Pless­ner, Betsch, Schal­lies, & Sch­wieren (2007). Au­to­matic on­line for­ma­tion of im­plicit at­ti­tudes to­ward poli­ti­ci­ans as a ba­sis for in­tu­itive vot­ing be­hav­ior. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 107-117). Psy­chol­ogy Press.

Raab & John­son (2007). Im­plicit learn­ing as a means to in­tu­itive de­ci­sion mak­ing in sports. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 119-133). Psy­chol­ogy Press.

Tver­sky & Kah­ne­man (1982). Judg­ment un­der un­cer­tainty: Heuris­tics and bi­ases. Cam­bridge Univer­sity Press.

Volz & Cra­mon (2007). Can neu­ro­science tell a story about in­tu­ition? In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 71-87). Psy­chol­ogy Press.

Zee­len­berg, Nelissen, & Pieters (2007). Emo­tion, mo­ti­va­tion, and de­ci­sion mak­ing: a feel­ing-is-for-do­ing ap­proach. In Pless­ner, Betsch, & Betsch (eds.), In­tu­ition in Judg­ment and De­ci­sion Mak­ing (pp. 173-189). Psy­chol­ogy Press.