Humans balance many goals while navigating social interactions. We are motivated not only by what we want for ourselves and for others, but also by what we want others to think about us. Inferences about agents’ personal and social goals have been formally captured using Bayesian inverse planning models. However, less is known about the mechanisms underlying our ability to infer presentational goals: an agent’s desires over how another agent sees them. We introduce a novel paradigm that allows us to test joint inferences about social and presentational goals. We propose an extension to inverse planning where agents derive additional utility from shifting others’ beliefs about their social goals. Because the model places no constraint on the valence of presentational targets, it naturally captures cases where agents wish to appear prosocial or adversarial. Across a variety of scenarios, participants make systematic joint inferences about social goals, presentational goals, and presentational targets. Our computational model captures participants’ inferences better than feature-based alternatives.
<< Back to list of publications