ECPR

Install the app

Install this application on your home screen for quick and easy access when you’re on the go.

Just tap Share then “Add to Home Screen”

ECPR

Install the app

Install this application on your home screen for quick and easy access when you’re on the go.

Just tap Share then “Add to Home Screen”

Generative AI and Opinion Formation: Evidence from a Survey Experiment on the British Electoral System

Political Psychology
Communication
Public Opinion
Survey Experiments
Technology
Rune Karlsen
Universitetet i Oslo
Rune Karlsen
Universitetet i Oslo

Abstract

The breakthrough of generative AI (GenAI) has sparked both optimism and concern about its societal impact. GenAI could disrupt how individuals are informed and form political beliefs, potentially intensifying polarization, misinformation, and political distrust in society. Alternatively, GenAI may democratize access to information, narrowing knowledge gaps and fostering better-informed citizens with well-informed opinions on vital societal issues (e.g. Jungherr & Schroeder, 2023; Schäfer, 2023). This paper explores the role of GenAI in opinion formation through a survey experiment embedded in a representative survey of Norwegian citizens. We examine the impact of GenAI (ChatGPT) on the formation, change, and reinforcement of attitudes toward the British electoral system (BES). As BES is recognizable but non-salient to most Norwegians, attitudes toward it are relatively malleable. The experiment compares media, ChatGPT vs. Wikipedia and online newspapers, and the tone of information provided by ChatGPT (neutral vs. politically biased, tailored for different education levels). We hypothesize that the impact of these treatments is conditioned by trust in GenAI, cognitive biases, and cognitive skills. Preliminary results indicate that one-third of respondents formed an opinion after exposure, with ChatGPT slightly less effective than Wikipedia in attitude formation. This underscores how people are quick to form an opinion based on relatively little information about an object, also from ChatGPT. Politically biased ChatGPT responses polarized attitudes, depending on the respondent’s ideology, emphasizing the role of AI in reinforcing political divides These findings underscore the dual-edged nature of GenAI, which, like prior technological breakthroughs, has the potential to both democratize information and deepen societal divides. In the recent past, rapid and consequential technological change– such as the rise of social media – democratized access to information and broke down old barriers to political participation, while simultaneously fueling developments in society that are currently driving us apart (Lorenz-Spreen et al., 2022). Similarly, this study's findings highlight the disruptive potential of AI and the need for continued research into how GenAI can counteract rather than exacerbate societal divisions.