Auditing OpenAI's GPT for hiring bias

As firms race to adopt generative AI to get an edge over competitors, reporters are designing experiments to measure the harms these technologies pose to consumers.
In this talk, investigative data journalist Leon Yin shares the background, methodology and findings for a recent data-driven investigation into hiring bias in OpenAI’s GPT. By reproducing a classic hiring discrimination study, Yin and his co-authors (Davey Alba and Leonardo Nicoletti) reveal GPT’s racial and gender biases against names when asked to rank equally-qualified candidates for the same role.