Race Distorts How Faces Seem (Un)trustworthy

What makes faces seem trustworthy versus untrustworthy? People often conflate facial trustworthiness with structural resemblances to specific emotions - angry-appearing faces seem untrustworthy, while happy-appearing faces seem trustworthy. Is this universal across the ethnicity of faces? Across two pre-registered studies, we built computational models of trustworthy appearance from perceivers' evaluations of Black, White, and East Asian synthetic faces (Study 1 Phase I: N = 405, 72% White; Study 2 Phase I: N = 436, 64% White). We manipulated new male faces to appear (un)trustworthy using the race-specific models, and raters assessed their expressions (Study 1 Phase II: N = 194, 68% White; Study 2 Phase II: N = 513, 70% White). For White targets, untrustworthiness judgments relied on structural resemblances with anger; for Black targets, on structural resemblances to fear and surprise; and for East Asian targets, on structural resemblances to multiple negatively valenced emotions. This inclusive approach broadens our understanding of how faces seem (un)trustworthy.