Abstract
<p>With the recent explosion of artificial intelligence (AI) in everyday life, questions have emerged about the abilities of AI to handle underrepresented groups. This study combines two recently developed resources—the Faces Expanded dataset and Python Facial Expression Analysis Toolbox (Py-Feat)—to compare human and AI emotion classification of images from underrepresented racial and age groups. The results showed a moderate positive correlation between human and AI performance, with humans outperforming AI in emotion classification. The relationship between human and AI performance was strongest for images displaying happiness and weakest for images displaying fear. These findings emphasize the importance of incorporating racial and age diversity into both facial stimuli and AI training datasets. This study contributes to the ongoing discussion about the limitations of AI in psychological research and its potential applications in real-world scenarios.</p>