Assessing Bias in Original & Updated Reading the Mind in the Eyes Test (RMET)

Presented by:
Heesu Kim, Harshini Raman, Talia Benheim, & Sophie Barowsky (Wellesley College)

The Reading the Mind in the Eyes Test (RMET) is a prevalent clinical measure for social cognition and theory of mind. However, due to the exclusive use of monochromatic pictures of white individuals and unintuitive vocabulary in the questionnaire, the original RMET has been updated to include diverse, full-color photographs of male and female faces as well as simpler vocabulary. In this study, we assess whether these revisions meaningfully address identified areas of potential bias in the original RMET. Using model selection, linear regression, measures of model fit, ANOVA, and post-hoc Tukey HSD, we found that ethnicity, gender, and native language were equally predictive in both the original and updated versions of the RMET, with non-European individuals consistently scoring the lowest. Our findings suggest that inherent qualities of the RMET, aside from choice of vocabulary and ethnicities of the photographed faces, may contribute to the biased predictivity of demographics on RMET performance.