Assessing Implicit Attitudes: What can be Learned from Simulations?

Abstract

Empirical studies involving the Implicit Association Test (IAT) have often revealed a general implicit preference for European-American as opposed to African-American stimuli. While it has been pointed out that this does not establish the existence of an implicit negative attitude toward the less preferred target concept, the existence (or absence) of such attitudes is empirically difficult to ascertain. We describe a computational model of performance on the Race-IAT through which the influence of attitudinal positivity or negativity on expected IAT performance is explored using simulations in which the strengths and nature of target-attribute associations in memory are manipulated. Results indicate that IAT effects readily emerge from different patterns of implicit associations without any need for absolutely positive or absolutely negative implicit attitudes. Pitting a simulation of the standard IAT against a simulated Sorting Paired Features Task demonstrates an advantage for the latter in distinguishing each of the implied target-attribute associations.