Joshua R. Gubler

Women often experience a deficit in persuasive influence, particularly in contexts like politics, where expectations about women's behavior may conflict with gendered expectations about the domain and role. In spite of this widespread pattern, the mechanisms behind the gender gap in persuasion are difficult to isolate and test in dynamic political conversations. Women may be less persuasive simply because role-based expectations prompt others to afford women less authority when speaking about politics generally or about some political topics. Alternatively, women may be less persuasive because they tend to use linguistic patterns or rhetorical strategies that are less effective. Using an original AI-based approach, we put 4,300 human subjects into realistic, real-time phone conversations with an AI agent. These agents were randomly assigned either a masculine or feminine gendered conversation topic and a male- or female-sounding voice, allowing us to manipulate topic and perceptions of speakers' gender while holding constant other elements of rhetorical content and strategy. After validating the experimental treatments, we find no direct effects of perceived gender of the AI agent, as cued by voice. However, gender affects other important aspects of the persuasive interaction. Research subjects were more likely to be persuaded about the masculine topic (defense spending) than the feminine topic (aid to the poor). Additionally, women research subjects were, on average, more persuaded than men. To the extent that men and women emphasize different topics in political conversations and have different discussion partners, these dynamics are likely to shape women's political persuasive influence.

Quantitative social science is in a new age, with many exciting possibilities driven by the development and widespread adoption of AI technologies. We articulate some of this promise here, but also point out important concerns for increasing AI integration. We use multiple cases of AI use in the social sciences to discuss important, scientific considerations that researchers should engage with to ensure that AI-empowered social science work contributes to broader scientific knowledge. We propose a version of AI infinite regress to highlight how ungrounded reliance on AI-centered social science production may create a vast body of work of unknown practical value. Ignoring these issues runs the risk of increased AI-enhanced output but decreased understanding and actual scientific and social benefit.