Think Better with Alexander Todorov

On Wednesday, March 9, Prof. Alexander Todorov presented his Think Better talk "What Do We 'See' in a Face?" exploring the science of first impressions. Todorov’s research explores how one of the most basic social stimuli, the face, impacts our judgment of that person. Faces are important to human development: primates evolved to seek out faces in particular. Todorov demonstrated our affinity to faces by projecting a face before the crowd: despite only seeing it for a fleeting 100 milliseconds, we can discern the person’s gender, age, race, and emotion. 

Beyond demographic attributes, we also make snap judgments about a person’s character just by their face: their trustworthiness, competence, aggressiveness, etc. While there is wide consensus in these judgments, are they reliable? Does a face truly give us insights into a person’s character?   

REAL WORLD IMPLICATIONS 

Snap judgments about faces influence all sort of assessments, from small everyday decisions to who we will vote for in an election. Just by studying politician’s faces, we can predict 70% of presidential election outcomes. Likewise, “competent-looking” CEOs are more likely to get higher compensation packages. Today, this is more than a heuristic tendency: some startups promise to profile personalities based on facial images alone, and some scientific publications claim that they can extrapolate criminal inclinations, sexual orientation, political leaning and so on from facial physiognomy. 

THE PROBLEM WITH JUDGING BASED ON FACES

While the physiognomic approach offers an easy solution to complex social judgements, it’s profoundly problematic. For instance, “the interview illusion” is a studied phenomenon which shows that interview impressions have little correlation with job performance; something like letters of recommendation offer much better indication of one’s job success. 

Some generalizations from facial judgments hold ground, but we can be too liberal about our extrapolations. In one study, images were taken of people before and after they didn’t sleep for 24 hours. Looking at these images alone, study participants were likely to call the sleep-deprived version of the person not as smart. On some level, this makes sense: a sleep deprived person will not cognitively perform as well if exhausted. However, inferring that the person was generally not intelligent does not hold the same ground. Another generalization that backfires is our association of competence with masculine features: when women have masculine features, we do not think of them as competent.

MAPPING OUR BIASES

Todorov’s research tries to capture the systematic biases in our judgements of people’s faces. Using the latest technology of deep neural networks, Todorov can model qualities like “trustworthiness” and “extroversion” on hyper-realistic, computer-generated faces. Of course, Todorov’s model is creating faces people will agree are more trustworthy or extroverted, but because these computer-generated people do not exist, these judgments do not correspond to any real character traits. 

WHY WE JUDGE BASED ON FACES

To summarize, first impressions based on facial appearance and other visual cues help us assess the intentions and capabilities of others, especially when good information is lacking (i.e. any time we’re interacting with strangers). But these shortcuts come at the expense of accuracy when considering the long-term characteristics of a person. Yes, a person who looks angry may be upset at the moment, but an isolated impression doesn’t prove they are typically an angry person. And judging a person’s character from a single frame of a photo is even shakier ground upon which to make a reliable judgment. 

UPCOMING EVENTS

Thank you to Professor Todorov for this fascinating presentation, to our moderator Nick Epley, and to the Gleacher Center for hosting this hybrid event.