Perception and Judgment Lab

Past and present research

Most of our work has been on characterizing the perceptual, cognitive, and neural basis of judgments from facial appearance. For a recent review of some of this work, see Todorov & Oh (2021). One of the important contributions of the lab was to introduce data-driven computational methods for visualizing the perceptual basis of social judgments from faces. These models capture systematic biases in judgments. See demonstrations for some of the resulting models of judgments. Recently, using generative adversarial networks we have built models capable of generating and manipulating the appearance of hyper-realistic faces. You can sign here to follow the development of the methods.

Present and future research I

How do stereotypes survive and are propagated even if they are not grounded in reality? How do these stereotypes lead to discrimination? When we think of bias and discrimination, we often think of direct discrimination, but much of discrimination is indirect. By choosing to interact with a “trustworthy-looking” partner, we can learn about their behavior, but not about the behavior of non-chosen partners. Moreover, these choices can systematically exclude non-preferred partners from mutually beneficial interactions. We are studying how our choices and features of the situation can lead to the confirmation and propagation of inaccurate stereotypes.

Present and future research II

To what extent are preferences idiosyncratic, rooted in our unique experience, or shared with others? We have developed methods for estimating idiosyncratic and shared contributions to preferences (Martinez, Funk, & Todorov, 2020). In the case of the only “pure” appearance judgment — attractiveness — about 50% of our preferences can be explained by idiosyncratic taste. For any other complex social judgment, more than 50% is explained by idiosyncratic taste. This phenomenon extends to all of our preferences. As a result, often what appears as “noise” in judgments is stable idiosyncratic variance. We are working on characterizing the reliability and the sources of our preferences.

Scroll to Top