Toward a Learning Account of Implicit Attitude Acquisition and Change

Abstract

The present dissertation seeks to characterize the learning processes by which implicit (indirectly revealed) evaluations of social and nonsocial stimuli are updated along three dimensions: (a) informational inputs to learning, (b) the nature of the learning process, and (c) the content and format of the resulting mental representations. Paper 1 provides evidence that implicit evaluations can be updated both as a result of direct experience with the environment (evaluative conditioning) and purely verbal information about upcoming stimulus pairings. Moreover, it shows that the two interventions are redundant in creating implicit attitude change, suggesting that they give rise to similar evaluative representations. Paper 2 offers further insight into the process by which evaluative conditioning shifts implicit evaluations, revealing that (a) learning can asymptote quickly, after as few as four stimulus pairings, and (b) information from stimulus pairings is productively combined with purely verbal information on the diagnosticity of those stimulus pairings. Furthermore, it demonstrates that direct experience with stimulus pairings creates more durable change in implicit evaluations than verbal descriptions of stimulus pairings. Finally, Paper 3 takes a reinforcement learning approach to show that whereas implicit evaluations are responsive to model-free learning, they are impervious to model-based learning. Overall, the current studies suggest that implicit attitudes can shift (a) as a result of certain (but not other) kinds of direct experience and certain (but not other) kinds of verbal interventions, (b) slowly or quickly, depending on the parameters of the task, but generally in a context-dependent manner, and (c) via creating highly compressed representations of value, be they associative (group A–good) or propositional (“group A is good”) in nature. As such, these results are partly compatible with propositional theories under which implicit attitudes can shift (a) in response to direct experience or language, (b) slowly or quickly, and (c) via creating propositional representations, but largely incompatible with associative theories under which implicit attitudes should shift exclusively (a) as a result of direct experience, (b) slowly, and (c) via creating associative representations. We discuss the theoretical and practical implications of these findings, along with avenues for future exploration that they open up.