Modeling the Influence of Cognitive Fluency and Stereotype Threat on the Processing of Implicit Attitudes

Modeling the Influence of Cognitive Fluency and Stereotype Threat on the Processing of Implicit Attitudes Boon-Kiat Quek (boonkiat.quek@northwestern.edu) Andrew Ortony (ortony@northwestern.edu) Department of Psychology, Northwestern University, Evanston, IL 60208, USA, and Institute of High Performance Computing, Agency for Science, Technology and Research, Singapore Abstract Studies reveal that the processing of implicit attitudes could be affected by individual differences in cognitive fluency, as well as by the presence of stereotype threat induced when subjects were primed with negative prejudices about their own social group. Using a previously proposed computational model of human performance on the Implicit Association Test, we examine possible processing mechanisms in which cognitive fluency and stereotype threat could influence the processing of implicit attitudes. Our goal is to extend the model to provide a cohesive and computationally plausible account for these effects; this is achieved by manipulating several model parameters that are analogous to human cogni- tive ability (in terms of processing speed and information retention ability) and shifts in confidence criteria for decision- making. Keywords: Implicit attitudes; cognitive ability; simulation; localist-connectionist networks. Introduction Implicit attitudes are generally assumed to underlie people’s thoughs, actions, choices and behavior ( Greenwald & Banaji, 1995). Understanding how such attitudes are processed might therefore provide some insight about why people behave in the way they do. Some ways in which such processes could be in- vestigated include affective priming ( Fazio, Sanbonmatsu, Powell & Kardes, 1986) and the Implicit Association Test (IAT; Greenwald, McGhee & Schwartz, 1998). The IAT was designed to assess automatic associations between con- cepts in memory. It relies on a simple two-choice response time paradigm which measures the time taken by subjects to classify sequentially presented input stimuli (words or im- ages) into one of two composite categories, each comprising a target concept (e.g., flower, insect) paired with an attribute concept (e.g., pleasant, unpleasant). Response latencies are expectedly shorter when targets are paired with compatible attributes (e.g., “flower or pleasant”, “insect or unpleas- ant”), and longer when paired with incompatible attributes (e.g., “flower or unpleasant”, “insect or pleasant”). The dif- ference in mean response times between compatible and incompatible categories is known as the IAT effect, and is taken as the relative preference for one target over another. Despite its wide application, many issues concerning the construct validity of the IAT have been raised (e.g., De Houwer, Teige-Mocigemba, Spruyt & Moors, 2009; Mierke & Klauer, 2003). Apart from automatic associations, per- formance on the IAT seems to also depend on various other factors, such as stimulus familiarity (Ottaway, Hayden & Oakes, 2001), concept saliency (Rothermund & Wentura, 2004), and extra-personal knowledge about prevailing cul- tural or societal norms (Karpinski & Hilton, 2001). Furthermore, several anomalous effects have also been ob- served. In a recent review, De Houwer, Teige-Mocigemba, Spruyt and Moors (2009) suggested that the processing of implicit attitudes could be influenced by differences in cog- nitive ability, citing McFarland and Crouch (2002) who observed significant correlations between response latencies and magnitudes of IAT effects, and Hummert, Garstka, O’Brien, Greenwald and Mellott (2002) who observed that IAT effects tended to increase with age. Given that process- ing speed is an important aspect of cognitive ability (Hunt, 1983) and declines with age (Salthouse, 1996), we will ex- pect subjects with lower cognitive abilities (especially with age-induced decline) to exhibit longer response latencies across all tasks on the IAT. Why this is associated with lar- ger IAT effects, however, remains to be determined. Another intriguing aspect of performance on the IAT is the possible role of stereotype threat. In a number of Race- IATs, Frantz, Cuddy, Burnett, Ray & Hart (2004) consis- tently observed that White subjects exhibited stronger pro- White IAT effects on the Race-IAT when they were in- structed beforehand that the test might expose their racial prejudices, as compared to other White subjects in control groups who were not similarly informed. Frantz et al. sug- gested that being told beforehand of the actual purpose of the Race-IAT would present a stereotype threat experience (Steele & Aronson, 1995) to the informed subjects, where knowledge of the test’s purpose might induce anxiety over the risk of confirming negative stereotypes about the racial attitudes that people in their social group are often presumed to endorse (e.g., being pro-White or anti-Black). Thus, we would expect subjects informed of the test’s purpose to have a greater interest in positive self-presentation and hence stronger motivation to respond in a more egalitarian manner (Frantz et al., 2004). Ironically, attempts to avoid the nega- tive stereotype appeared to interfere with performance on the Race-IAT, producing a stronger pro-White IAT effect instead of reducing it. However, no suggestions were pro- vided to explain how such task interference might have taken place, nor the manner in which strategies for coping with the stereotype threat experience might have backfired. Both the cognitive fluency effect and the stereotype threat effect are noteworthy because they have important implica- tions for our understanding of the nature of information processing that underlie performance on the IAT. In this paper, we examine some of these implications, and propose