The Effect of Patient Gender on ChatGPT’s Success When Answering Medical MCQs

• In Event: Electronic Board Session IV

Abstract

One concern about AI systems is their potential to reproduce biases present in their training data. In this paper, we investigate one possible manifestation of such bias by asking ChatGPT to answer gender-neutral medical MCQs rewritten to include patient gender. Results suggest that gender affects ChatGPT’s success on these items.

Authors

• Peter Baldwin, NATIONAL BOARD OF MEDICAL EXAM

• Nina James Braum-Bharti, Germantown Friends School

• Miles Somay Baldwin, Harriton