People reliably and immediately make intuitive judgments of other people’s traits (e.g., intelligence, kindness) from their appearances, irrespective of the accuracy of the judgments. Many of these judgments influence important social outcomes, such as voting, hiring, and mating (Todorov et al. 2015). My research focuses on understanding the mechanisms underlying these intuitive judgments.

Data-Driven Modeling of Social Perception

Data-driven modeling of social impressions (Jack & Schyns 2017) can identify the visual information responsible for these impressions (e.g., an impression of competence) with little prior assumptions as to what facial features matter. Some of the features that matter reflect stereotypes prevalent in society. In my recent work (Oh, Buck, & Todorov, in press), I showed that masculinity is one of the main ingredients of impressions of competence. This gender stereotype was not apparent because competence impressions are strongly correlated with facial attractiveness, which tends to be negatively correlated with masculinity. Taking advantage of the computational nature of the models of impressions, we built a model of competence that can manipulate perceived competence while controlling for attractiveness. This model removes the strong halo effect on competence impressions and shows that competent-looking faces are more masculine. This line of research not only identifies the visual information that affects social impressions but also uncovers hidden stereotype biases in these impressions.

Screen Shot 2017-11-23 at 10.29.52 AMScreen Shot 2017-11-23 at 10.01.22 AM
Controlling for facial attractiveness, a natural confound of competence impressions, more competent faces look more masculine (x-axis = the manipulation level of the competence impression models; y-axis = to what extent people categorized faces as male; arrays of faces = two models used to manipulate faces on their perceived competence).

Variance in Social Perception

Both characteristics of target persons and perceivers shape the process of impression formation. Among these characteristics, target individuals’ gender and perceivers’ expectations about gender (e.g., gender stereotypes) play a key role. In recent work, using dimensionality reduction and data-driven face modeling, I found that when targets are women, people’s impressions of the targets are more simplified (e.g., higher correlations among the impressions of warmth, dominance, attractiveness, happiness, emotional stability, and so on) and are more highly tied to overall positivity/negativity than when the targets are men (Oh, Dotsch, Porter, & Todorov, under review). Further, when perceivers are more willing to endorse gender stereotypes, they show more simplified impressions of women. This line of research informs us about the importance of social categories and individual differences in social perception.

Screen Shot 2017-11-22 at 5.22.52 PM
People tend to have less differentiated impressions of women than of men (y-axis represents the level of correlations among impressions; each dot denotes the level of correlations between a pair of impression ratings).

Neural Responses to Social Stimuli

Data-driven modeling (Jack & Schyns 2017) can be used to identify not only the visual information underlying specific impressions but also the visual information that drives both perceptual and neural responses to social stimuli. Shuo Wang and I analyzed electrocorticographic (ECoG) responses to faces and showed that different types of social information (e.g., race, age, social impressions) are represented in different brain areas in different stages of processing. This line of research informs us about the processing of socially salient information over time and the prioritization of visual information in the visual system.


Alexander Todorov (Princeton University)
Ron Dotsch (Philips Research)
Eldar Shafir (Princeton University)
Daniel Osherson (Princeton University)
Shuo Wang (West Virginia University)
Joel E. Martinez (Princeton University)
Alex Koch
(University of Cologne)
Evan W. Carr (Exponent Human Factor)