Humans vs. AI or Women vs. AI? Gender, Task Type, and Influence

• In Event: AI and the Value of Work

Abstract

People are increasingly collaborating with artificial intelligence tools that provide advice, feedback, or recommendations. Yet little is known about whether the social influence processes that reproduce gender inequality among people extend to human-AI interaction. Building on status characteristics and expectations states theory, which explains how gendered status beliefs shape influence hierarchies, we ask whether the gender presentation of AI tools affects people’s willingness to accept their influence. We report findings from two experiments. In Study 1, participants collaborated with either a human or AI partner, presented as male or female, in a decision-making task. Consistent with prior sociological research, participants accepted less influence from women than from men. However, this inequality did not extend to AI: participants accepted as much influence from female AI partners as they did from male AI and from men. In Study 2, we tested the robustness of the null effect for AI gender across two different task types. Again, we found that the gender inequalities in influence observed in people do not extend to AI, regardless of the task. Our results suggest that while gender continues to shape influence among humans, AI operates outside traditional status hierarchies—reshaping, but not erasing, gendered patterns of influence.

Authors

• Malissa Alinor, University of North Carolina-Chapel Hill

• Ashley Harrell, Duke University