Is implicit cognition bad cognition?


by Sophie Stammers– incom­ing postdoc­tor­al fel­low on pro­ject PERFECT

A sig­ni­fic­ant body of research in cog­nit­ive sci­ence holds that human cog­ni­tion com­prises two kinds of pro­cesses: expli­cit and impli­cit. According to this research, expli­cit pro­cesses oper­ate slowly, requir­ing atten­tion­al guid­ance, whilst impli­cit pro­cesses oper­ate quickly, auto­mat­ic­ally and without atten­tion­al guid­ance (Kahneman, 2012; Gawonski and Bodenhausen, 2014). A prom­in­ent example of impli­cit cog­ni­tion that has seen much recent dis­cus­sion in philo­sophy is that of impli­cit social bias, where asso­ci­ations between (often) stig­mat­ized social groups and (often) neg­at­ive traits mani­fest in beha­viour, res­ult­ing in dis­crim­in­a­tion (see Brownstein and Saul, 2016a; 2016b). This is the case even though the indi­vidu­al in ques­tion isn’t dir­ect­ing their beha­viour to be dis­crim­in­at­ory with the use of atten­tion­al guid­ance, and is appar­ently unaware that they’re exhib­it­ing any kind of dis­fa­vour­ing treat­ment at the time (although see Holroyd 2015 for the sug­ges­tion that indi­vidu­als may be able to observe bias in their beha­viour).

Examples of impli­cit social bias mani­fest­ing in beha­viour include exhib­it­ing great­er signs of social unease, less smil­ing and more speech errors when con­vers­ing with a black exper­i­menter com­pared to when the exper­i­menter is white (McConnell and Leibold, 2001); less eye con­tact and increased blink­ing in con­ver­sa­tions with a black exper­i­menter versus their white coun­ter­part (Dovidio et al., 1997), and reduced will­ing­ness for skin con­tact with a black exper­i­menter versus a white one (Wilson et al., 2000). Implicit social biases also arise in more delib­er­at­ive scen­ari­os: Swedish recruit­ers who har­bor impli­cit racial asso­ci­ations are less likely to inter­view applic­ants per­ceived to be Muslim, as com­pared to applic­ants with a Swedish name (Rooth, 2007), and doc­tors who har­bor impli­cit racial asso­ci­ations are less likely to offer treat­ment to black patients with the clin­ic­al present­a­tion of heart dis­ease than to white patients with the same clin­ic­al present­a­tion of the dis­ease (Green, et al., 2007). These stud­ies estab­lish that there is no cor­rel­a­tion between par­ti­cipants’ dis­crim­in­at­ory beha­viour and the beliefs and val­ues that they pro­fess to have when ques­tioned.

Both the mech­an­isms of impli­cit bias, and impli­cit pro­cesses more gen­er­ally, are often char­ac­ter­ised in the lan­guage of the sup-optimal. Variously, they deliv­er “a more inflex­ible form of think­ing” than expli­cit cog­ni­tion (Pérez, 2016: 28), they are “ara­tion­al” com­pared to the ration­al pro­cesses that gov­ern belief update (Gendler, 2008a: 641; 2008b: 557), and their con­tent is “dis­uni­fied” with our set of expli­cit atti­tudes (Levy, 2014: 101–103). As such, one might be temp­ted to think of impli­cit cog­ni­tion as reg­u­larly, or even neces­sar­ily bad cog­ni­tion. A strong inter­pret­a­tion of that value-laden assess­ment might mean that the pro­cesses in ques­tion deliv­er object­ively bad out­puts, how­ever these are to be defined, but we could also mean some­thing a bit weak­er, such as that out­puts are not aligned with the agent’s goals, or sim­il­ar. It’s easy to see why one might apply this value-laden assess­ment to the mech­an­isms which res­ult in impli­citly biased beha­viour: indi­vidu­als simply have no reas­on to dis­crim­in­ate against already mar­gin­al­ized people in the ways out­lined above, and yet they do any­way – that seems like a good can­did­ate for bad cog­ni­tion. That impli­citly biased beha­viours are the product of what appears to be a sub­op­tim­al pro­cessing sys­tem might motiv­ate the argu­ment that we’re not the agents of our impli­citly biased beha­vi­ors, as well as argu­ments that might fol­low from this, such as that it is not appro­pri­ate to hold people mor­ally respons­ible for their impli­cit biases (Levy, 2014).

But I think it would be wrong to con­clude that impli­cit cog­ni­tion neces­sar­ily deliv­ers sub­op­tim­al out­puts, and that impli­cit bias is an example of bad cog­ni­tion simply for the reas­on that it is impli­cit. Moreover, as I’ll argue below, main­tain­ing the former claim may well do a dis­ser­vice to the pro­ject of redu­cing impli­cit social biases.

Whilst expli­cit pro­cesses may be ‘bet­ter’ at some cog­nit­ive tasks, research sug­gests that impli­cit pro­cesses can actu­ally deliv­er a more favour­able per­form­ance than expli­cit pro­cesses in a vari­ety of domains. For instance, non-attentional, auto­mat­ic pro­cesses gov­ern the fast motor reac­tions employed by skilled ath­letes (Kibele, 2006). Trying to bring these pro­cesses under atten­tion­al con­trol can actu­ally dis­rupt sport­ing per­form­ance: Fleagal and Anderson (2008) show that dir­ect­ing atten­tion to their action per­form­ance sig­ni­fic­antly impairs the abil­ity of high-skill golfers on a put­ting task, whilst high-skill foot­ballers per­form less pro­fi­ciently when dir­ect­ing atten­tion to their exe­cu­tion of drib­bling (Beilock et al., 2002). Engaging atten­tion­al pro­cesses when learn­ing new motor skills can also dis­rupt per­form­ance (McKay et al., 2015).

Meanwhile, func­tion­al MRI stud­ies sug­gest that impro­visa­tion implic­ates non-attentional pro­cesses. One study shows that when pro­fes­sion­al jazz pian­ists impro­vise, they do so in the absence of cent­ral pro­cesses implic­ated in atten­tion­al guid­ance (Limb and Braun, 2008). Another study demon­strates that trained musi­cians inhib­it net­works asso­ci­ated with atten­tion­al pro­cessing dur­ing impro­visa­tion, (Berkowitz and Ansari, 2010).

Further, delib­er­ately dis­en­ga­ging atten­tion­al resources can facil­it­ate cre­ativ­ity, a pro­cess known as ‘incub­a­tion’. Subjects who return to work on a cre­at­ive task after a peri­od dir­ect­ing atten­tion­al resources to some­thing unre­lated to the task at hand often deliv­er enhanced out­puts com­pared with those who con­tinu­ally engage their atten­tion­al resources (Dodds et al., 2003). It has been pro­posed that task-relevant impli­cit pro­cesses remain act­ive dur­ing the incub­a­tion peri­od and con­trib­ute to enhanced cre­at­ive out­put (Ritter and Dijksterhuis, 2014).

So it would be wrong to sug­gest that impli­cit pro­cesses neces­sar­ily, or even typ­ic­ally, deliv­er sub-optimal out­puts com­pared with their expli­cit cous­ins. And per­tin­ent to our dis­cus­sion of impli­cit social bias, impli­cit pro­cesses them­selves can actu­ally be recruited to inhib­it the mani­fest­a­tion of bias. Research demon­strates that par­ti­cipants with genu­ine long-term egal­it­ari­an com­mit­ments (Moskowitz et al. 1999) as well as those in whom egal­it­ari­an com­mit­ments are activ­ated dur­ing in an exper­i­ment­al task (Moskowitz and Li, 2011) actu­ally mani­fest less impli­cit bias than those without such com­mit­ments. Crucially, the pro­cesses which bring impli­cit responses in line with an agent’s long-term com­mit­ments are not driv­en by atten­tion­al guid­ance, instead oper­at­ing auto­mat­ic­ally to pre­vent the facil­it­a­tion of ste­reo­typ­ic cat­egor­ies in the pres­ence of the rel­ev­ant social con­cepts (Moskowitz et al. 1999: 168). The sug­ges­tion here is that devel­op­ing genu­ine com­mit­ments to egal­it­ari­an val­ues and treat­ment can actu­ally recal­ib­rate impli­cit pro­cesses to deliv­er value-consistent beha­vi­or (see Holroyd and Kelly, 2016), without need­ing to effort­fully over­ride impli­cit responses each time one encoun­ters social con­cepts that might oth­er­wise trig­ger biased reac­tions. It would seem that the pro­file of impli­cit pro­cesses as inflex­ible, ara­tion­al and dis­uni­fied with expli­cit val­ues and com­mit­ments is ill-fitted to account for this example.

So, in a num­ber of cases it seems that impli­cit pro­cesses can serve our goals and val­ues. If this is right, then we should per­haps be more will­ing to loc­ate ourselves as agents not just in the beha­vi­or that arises from our expli­cit pro­cesses, but in that which arises from our impli­cit ones as well.

I think this has an import­ant implic­a­tion for prac­tices related to impli­cit bias train­ing. We should be wary of the rhet­or­ic that dis­tances us as agents from our impli­cit pro­cesses: for instance, char­ac­ter­iz­ing impli­cit bias as “racism without racists”1 might be com­fort­ing for those of us with impli­cit racial biases, but dis­own­ing the impli­cit pro­cesses that lead to racial dis­crim­in­a­tion, while not dis­own­ing those that lead to skilled music­al impro­visa­tion or cre­ativ­ity as above, seems some­what incon­sist­ent. I won­der wheth­er great­er will­ing­ness to accept one’s impli­cit pro­cesses as aspects of one’s agency (not neces­sar­ily as cent­ral, defin­ing aspects of one’s agency — but some­where in there non­ethe­less) might help to motiv­ate the pro­ject of realign­ing one’s impli­citly biased responses?



  1. In U.S. Department of Justice. 2016. “Implicit Bias.” Community Oriented Policing Services report, page 1. Accessed 27/07/16, URL:



Berkowitz, A. L. and D. Ansari. 2010. “Expertise-Related Deactivation of the Right Temporoparietal Junction dur­ing Musical Improvisation.” NeuroImage 49 (1): 712–19.

Brownstein, M and J. Saul. 2016a. Implicit Bias and Philosophy, Volume 1: Metaphysics and Epistemology, New York: Oxford University Press.

Brownstein, M and J. Saul. 2016b. Implicit Bias and Philosophy, Volume 2: Moral Responsibility, Structural Injustice, and Ethics, New York: Oxford University Press.

Dodds R. D., T. B. Ward and S. M. Smith. 2003. “Incubation in prob­lem solv­ing and cre­ativ­ity.” in The Creativity Research Handbook, edited by Runco M. A., Cresskill, NJ: Hampton Press.

Dovidio, J. F., K. Kawakami, C. Johnson, B. Johnson and A. Howard. 1997. “On the Nature of Prejudice: Automatic and Controlled Processes.” Journal of Experimental Social Psychology 33 (5): 510–40.

Gawronski, B. and G. V. Bodenhausen. 2014. “Implicit and Explicit Evaluation: A Brief Review of the Associative-Propositional Evaluation Model: APE Model.” Social and Personality Psychology Compass 8 (8): 448–62.

Gendler, T. S. 2008a. “Alief and Belief.” The Journal of Philosophy 105 (10): 634–63.

———. 2008b. “Alief in Action (and Reaction).” Mind & Language 23 (5): 552– 85.

Green, A. R., D. R. Carney, D. J. Pallin, L. H. Ngo, K. L. Raymond, L. I. Iezzoni and M. R. Banaji. 2007. “Implicit Bias among Physicians and Its Prediction of Thrombolysis Decisions for Black and White Patients.” Journal of General Internal Medicine 22 (9): 1231–38.

Holroyd, J. 2015. “Implicit Bias, Awareness and Imperfect Cognitions.” Consciousness and Cognition 33 (May): 511–23.

Holroyd, J. and D. Kelly. 2016. “Implicit Bias, Character, and Control.” In From Personality to Virtue, edited by A. Masala and J. Webber, Oxford: Oxford University Press.

Kahneman, D. 2012. Thinking, Fast and Slow, London: Penguin Books.

Kibele, A. 2006. “Non-Consciously Controlled Decision Making for Fast Motor Reactions in sports—A Priming Approach for Motor Responses to Non-Consciously Perceived Movement Features.” Psychology of Sport and Exercise 7 (6): 591–610.

Levy, N. 2014. Consciousness and Moral Responsibility, Oxford; New York: Oxford University Press.

Limb, C. J. and A. R. Braun. 2008. “Neural Substrates of Spontaneous Musical Performance: An fMRI Study of Jazz Improvisation.” Edited by E. Greene. PLoS ONE 3 (2): e1679.

McConnell, A. R. and J. M. Leibold. 2001. “Relations among the Implicit Association Test, Discriminatory Behavior, and Explicit Measures of Racial Attitudes.” Journal of Experimental Social Psychology 37 (5): 435–42.

McKay, B., G. Wulf, R. Lewthwaite and A. Nordin. 2015. “The Self: Your Own Worst Enemy? A Test of the Self-Invoking Trigger Hypothesis.” The Quarterly Journal of Experimental Psychology 68 (9): 1910–19.

Moskowitz, G. B., P. M. Gollwitzer, W. Wasel and B. Schaal. 1999. “Preconscious Control of Stereotype Activation Through Chronic Egalitarian Goals.” Journal of Personality and Social Psychology 77 (1): 167–184

Moskowitz, G. B., and P. Li. 2011. “Egalitarian Goals Trigger Stereotype Inhibition: A Proactive Form of Stereotype Control.” Journal of Experimental Social Psychology 47 (1): 103–16.

Pérez, E. O. 2016. Unspoken Politics: Implicit Attitudes and Political Thinking, New York, NY: Cambridge University Press.

Ritter, S. M. and A. Dijksterhuis. 2014. “Creativity–the Unconscious Foundations of the Incubation Period.” Frontiers in Human Neuroscience 8: 22–31.

Rooth, D‑O. 2007. “Implicit Discrimination in Hiring: Real World Evidence.” (IZA Discussion Paper No. 2764). Bonn, Germany: Forschungsinstitut Zur Zukunft Der Arbeit (Institute for the Study of Labor).

Wilson, T. D., S. Lindsey and T. Y. Schooler. 2000. “A Model of Dual Attitudes.” Psychological Review 107 (1): 101–26.