In the APSA Public Scholarship Program, graduate students in political science produce summaries of new research in the American Political Science Review. This piece, written by Jack Wippell, covers the new article by Yamil Ricardo Velez and Patrick Liu, “Confronting Core Issues: A Critical Assessment of Attitude Polarization Using Tailored Experiments.”
Do opposing arguments really change minds, or do they harden beliefs? In their recent APSR article, Yamil Ricardo Velez and Patrick Liu challenge the idea that exposure to opposing views always either reduces or increases divisions. Instead, their findings reveal that it’s not just what we argue about, but how we argue that matters. Through experiments using cutting-edge GPT-3 AI technology to customize counterarguments, the authors show that tone and intensity, especially when crossing into hostility, are the true drivers of polarization. Polarization might not be inevitable—it’s simply waiting for the right kind push.
Velez and Liu begin by discussing prior research on how people form political beliefs and how correcting misinformation can sometimes have the opposite effect, making people hold onto false beliefs even more strongly. However, they also point out that recent studies have shown mixed results, as people do sometimes update their beliefs when presented with alternative information. In particular, the authors suggest that these mixed findings may be due to the failure of existing experiments to target attitudes that are deeply held and personally relevant to participants.
In response, the authors propose a novel approach: measuring participants’ core issue positions and exposing them to personalized, emotionally charged counter-attitudinal information that is automatically generated by the large language model GPT-3. Across a series of experiments, the authors ask participants to provide open-ended responses about the political issues they care most deeply about. GPT-3 then generates a series of counter-attitudinal arguments based on these responses, allowing for highly personalized treatments.
“Their innovative use of AI to generate tailored arguments offers a fresh methodological approach to the study of political attitudes, while their findings carry significant implications for understanding how the tone and emotional content of political arguments can drive polarization.”The first finding that emerges is that attitude polarization is not as common as some scholars have previously suggested. Across their first two experiments, participants exposed to neutral or mildly counter-attitudinal arguments did not become more entrenched in their beliefs; instead, they either moderated their positions or showed no significant change in attitude strength or certainty. These results challenge assumptions that exposure to opposing viewpoints inevitably leads to polarization.
The story becomes more complex when the authors introduce emotionally charged arguments. In their third experiment, Velez and Liu intensify the treatment by generating first-person, paragraph-length counter-attitudinal arguments. Once again, they fail to find evidence of attitude polarization, with participants showing modest signs of moderation. However, in their final two experiments, where the arguments are not only emotionally charged but also vitriolic and uncivil, the authors detect clear evidence of polarization.
These findings could not be timelier in light of the increasing political polarization and toxic public discourse we observe across the United States. Social media platforms, like X and Facebook, where contentious and hostile political discussions proliferate, may be creating the exact conditions that Velez and Liu describe as triggers for polarization: emotionally charged, vitriolic arguments. As public debates grow more heated, this research suggests that the nature of these debates—rather than simply their mere existence—may be the factors fueling division.
In sum, Velez and Liu’s study provides a sophisticated and timely examination of attitude polarization in today’s hyperpolarized political environment. Their innovative use of AI to generate tailored arguments offers a fresh methodological approach to the study of political attitudes, while their findings carry significant implications for understanding how the tone and emotional content of political arguments can drive polarization. As political rhetoric continues to intensify in the digital age, their work offers valuable insights for policymakers, media platforms, and researchers alike.
- Jack Wippell is a PhD Student in the Department of Sociology at The Ohio State University. His research interests cover political sociology, social movements, and culture, and his current focus is on the emergence, spread and mobilization of far-right extremism across the United States and Europe. He also has interests at the intersection of computational and qualitative methodologies.
- VELEZ, YAMIL RICARDO, and PATRICK LIU. 2024. “Confronting Core Issues: A Critical Assessment of Attitude Polarization Using Tailored Experiments.” . American Political Science Review, 1–18.
- About the APSA Public Scholarship Program.
Through the use of movies, documentaries, quotes, and documents, we typically have an interpretation of colonial political debates and exchanges as being “heated.” Thanks to this recent study, we can now assume that polarization happened in those times.
What is interesting about this is that polarization is typically viewed negatively; yet, with polarization in colonial times came the establishment of a government/new country the likes of which were never seen in human history, with prosperity and freedom the likes of which were never seen in human history, and has evolved into a world super-power.