Abstract

Individual implicit biases (automatic, typically uncontrollable, mainly unconscious attitudes towards social groups, which can be measured by, e.g., the IAT) have been shown to vary over time and to be poor predictors of discriminating behaviour. On the other hand, collective levels of implicit biases (e.g., average IAT-scores on regional or national levels) turn out to be quite stable, as well as quite good predictors of social inequities and vulnerabilities (e.g., unequal educational achievements, or health outcomes).

The Bias of Crowds paradigm aims to reconcile these findings by analysing implicit bias as a collective phenomenon which reflects the accessibility of concepts that are linked to social groups (e.g., stereotypes). While this accessibility, on an individual level, varies considerably over time and context, on a collective level, such idiosyncratic random variations cancel out, such that the aggregate reveals an underlying systemic bias against certain social groups.

This characterisation of the Bias of Crowds paradigm suggests a close connection to the Wisdom of Crowds paradigm, according to which aggregating judgments across individuals reduces error variance and reveals meaningful information in means. This paper explores the connection between the two paradigms, with an eye to what they can teach us if we want to overcome the pernicious effects of implicit biases. Under which conditions does a crowd’s aggregate bias turn into wisdom? And how can we translate these conditions into practical procedures for, e.g., hiring or health-care allocating decisions, which are prone to discrimination?