The Second Order Effects of LLMs and Generative AI on Political Communication

• In Event: AI and the Future of Democracy

Abstract

How will generative AI systems and large language models (LLMs) like ChatGPT impact political participation and communication? Prior work shows that legislators largely cannot distinguish between machine-generated and human-generated communication (Kreps and Kriner 2023). Moving beyond these persuasive impacts of machine-generated messages, we seek to explore broader second order effects of generative AI, examining its impact on political attitudes, communication, and behavior. For example, does awareness of the possible use of generative AI to draft individual emails or even to facilitate systematic use in large advocacy campaigns change the way that members of the public think about political participation? Does it alter how the public feels regarding whether their communication is valued by their elected political representatives, or their intention to reach out to politicians? To study this, we administer an online survey experiment to American adults that randomly varies the presentation of information about how LLMs like ChatGPT can be used in political communication. We examine the impacts of generative AI on the perceived effectiveness of and support for using email or other communication strategies for contact with representatives, among other outcomes.

Authors

• Bryce Dietrich, Purdue University

• Kaylyn Jackson Schiff, Purdue University

• Daniel S. Schiff, Purdue University