Implicit Bias as Automatic Behavior

Abstract

Researchers interested in implicit bias agree that no one agrees what implicit bias is. Gawronski, Ledgerwood, and Eastwick (this issue) join a spate of scholars calling for better conceptual clarity around what it means for a construct or a measure to be implicit (Corneille & H€ utter, 2020; Fazio, Granados Samatoa, Boggs, & Ladanyi, 2022; Schmader, Dennehy, & Baron, 2022; Van Dessel et al., 2020). Some argue we should do away with the term entirely (Corneille & H€ utter, 2020), and others argue that authors simply need to do a better job defining how they are idiosyncratically using the term each time they use it (Greenwald & Lai, 2020). In their target article, Gawronski et al. argue for a fundamental redefinition of what it means for bias to be implicit. More specifically, they argue that implicit bias (IB) and bias on implicit measures (BIM) are conceptually and empirically distinct, and that BIM (defined as “effects of social category membership on behavioral responses captured by measurement instruments conventionally describe as implicit”) should not be treated as an instance of IB (defined as “behavioral responses influenced by social category cues when respondents are unaware of the effect of social category cues on their behavioral responses”). We agree that the time has come for our definition of implicit to be revamped in light of new findings. In fact, it is past time; we co-chaired a symposium titled “What is implicit about implicit attitudes?” at the Society for Personality and Social Psychology’s annual meeting in 2009, more than a decade ago. And we applaud the authors of the target article for taking a bold step toward making a change. Further, we agree with them that bias is best defined as a behavioral phenomenon rather than a latent mental construct. This is not a statement we make lightly; it has required some serious scholarly contemplation of the current state of the literature and some serious non-scholarly contemplation of our own egos to reach this conclusion. For some time now we, like most others, have described implicit bias as something that people have–e.g., participants have an implicit bias favoring one novel individual over another (Ratliff & Nosek, 2011), have an implicit preference favoring White over Black Americans (Chen & Ratliff, 2018), or have an implicit positive or negative attitude toward feminists (Redford, Howell, Meijs, & Ratliff, 2018). Many of us are quite invested in this way of thinking. And change is hard! But we recognize that we gain a lot by taking this more functional approach to bias. Most notably, a functional approach allows researchers to circumvent the perplexing situation of using the same name for construct and measure. Further, many of us working in this area are doing so because we hope to provide insights through which people can change their behavior in order to reduce inequality on real life issues that matter. Given that the problem of bias is a behavioral problem (De Houwer, 2019), it makes sense to define bias in behavioral terms. So let us agree to define bias as the influence of social category cues on behavioral responses. But we are still left, however, with the problem of what makes bias implicit. To that end, we would like to raise two concerns about the target article. First, if the authors had proposed that BIM should not necessarily be treated as an instance of IB, we would concur; but we disagree with the strong language implying that BIM should never be considered an instance of IB; “does not equal” is not the same as “is orthogonal to.” Second, we do not agree that awareness (which the original authors use interchangeably with consciousness) is the only or best factor by which to distinguish implicit from explicit bias. Consciousness is messy business, and it is nearly impossible to delineate whether any given effect is unconscious or conscious as most, maybe all, have aspects of both. We would instead argue for distinguishing between implicit and explicit bias based on features of automaticity (Moors & De Houwer, 2006).