Psychologists and philosophers have not yet resolved whatthey take implicit attitudes to be; and, some, concerned aboutlimitations in the psychometric evidence, have evenchallenged the predictive and theoretical value of positingimplicit attitudes in explanations for social behavior. In themidst of this debate, prominent stakeholders in science havecalled for scientific communities to recognize andcountenance implicit bias in STEM fields. In this paper, Istake out a stakeholder conception of implicit bias thatresponds to these challenges in ways that are responsive to thepsychometric evidence, while also being resilient to the sortsof disagreements and scientific progress that would notundermine the soundness of this call. Along the way, myaccount advocates for attributing collective (group-level)implicit attitudes rather than individual-level implicitattitudes. This position raises new puzzles for future researchon the relationship (metaphysical, epistemic, and ethical)between collective implicit attitudes and individual-levelattitudes.

I motivate and articulate a dispositional account of aversiveracism. By conceptualizing and measuring attitudes in termsof their full distribution, rather than in terms of their mode ormean preference, my account of dispositional attitudes givesambivalent attitudes (qua attitude) the ability to predictaggregate behavior. This account can be distinguished fromother dispositional accounts of attitude by its ability tocharacterize ambivalent attitudes such as aversive racism atthe attitudinal rather than the sub-attitudinal level and itsdeeper appreciation of the analogy between traits andattitudes.