Accuracy and bias may seem at odds in perception. Yet they are as much complementary as they are contradictory. For example, people are excellent at recognizing others’ expressions of emotion, but when faces are hard to see, people are biased to report that individuals look threatening, or angry. Error in name only, these sorts of negative biases may be protective or functional, and they can be found alongside countless examinations of accuracy in the vision and social cognition literature. Yet little work in the field of ensemble coding has examined bias during the perception of crowds. This is surprising because biases, especially negative biases, should manifest differently for judgments about many people seen at once, or in succession. Crowds exert different pressures on perceivers than do individuals, people behave differently when in crowds than they do when they’re alone, and people hold different beliefs about crowds than individuals. In this talk, I will share new findings demonstrating that bias to report anger is greatly amplified when people make judgments about crowds compared with individuals. Furthermore, perceivers are biased in the way they attend to faces in crowds, which leads them to estimate that crowds are more emotional than they actually are, especially when those crowds are angry. Finally, I will share plans for a new program of work (including drift-diffusion modeling, eye-tracking, and Generative Adversarial Networks) to examine how the identities of people in a crowd (e.g., race, gender, age, etc.) impact biased judgments of their emotion. https://alumni.du.edu/about/faculty-directory/timothy-daniel-sweeny