Humans effortlessly reason about others’ emotions all the time. However, scientists do not have formal, computational models to describe how humans actually do this reasoning. New work by Desmond Ong, Jamil Zaki and Noah Goodman proposes that humans might do such reasoning rationally. Across four studies, they show that a rational Bayesian model of emotional reasoning accurately describes human inferences. In particular, the paper tackles the problem of emotional cue integration (see also Professor Zaki's 2013 paper in Perspectives on Psychological Science). Imagine seeing a friend’s smiling face when you know that something bad happened: how do you reconcile whether your friend is feeling positive or negative? This problem involves combining the potentially conflicting information you get from multiple cues (smiling face; bad outcome). We show that, just like other cue integration problems in psychology (e.g. using visual and audio cues), we can model human inferences of emotion using an optimal Bayesian model. In addition to informing basic psychological theory, this work has many exciting applications like applying computational models to mental health (computational psychiatry) or affective computing. To read more publications from the SSNL, click here.