Talking politics with AI: How chatbots can change our political opinions and decisions

Being rapidly adopted by millions of users worldwide, AI chatbots are increasingly serving as alternatives to search engines and transforming information-seeking practices. Yet, such tools are prone to intrinsic biases (for instance, based on gender or race) and errors, often referred to as hallucinations. Furthermore, due to their high levels of anthropomorphism, personalization, and sensitivity to human feedback, they can be highly persuasive and may reinforce individual beliefs. Drawing on empirical studies on the use of generative AI, we will talk about how AI chatbots can influence the way we learn about politics and what an increased reliance on AI may mean for democracy. In particular, we will discuss whether users are ready to ask AI for political advice, how they formulate prompts on political topics, whether AI chatbots are more trusted than search engines, and how the extensive personalization of AI-generated outputs may shape users’ perceptions of social reality.

Dieser Programmpunkt wird durch die Stiftung Mercator unterstützt. / This programme session is supported by Stiftung Mercator.