Beyond Belief Change: LLM-Generated Counterargument Experiments on Political Attitudes

American Political Science Review Dataverse hosts replication data for a study on political persuasion and belief relevance. The research involved experiments with two large online convenience samples, using large language models to generate counterarguments targeting specific beliefs. Yamil Velez authored the dataset, which was last updated on June 18, 2026.

Use Cases

• Analyzing the durability of attitude change based on the 10-day follow-up measurement mentioned in the description
• Comparing the effect sizes of counterarguments targeting focal versus distal beliefs as described in the experimental design
• Studying the relationship between belief strength and political attitude formation using the elicited belief data
• Replicating experimental results on the persuasive returns of targeting attitude-relevant beliefs

Strengths

• Data supports a study published in the American Political Science Review, suggesting peer-reviewed rigor
• Experiments were conducted with two large online convenience samples, indicating a substantial participant pool
• Includes longitudinal measurement with effects tracked after 10 days

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Data may reflect bias inherent to online convenience samples

Provenance

Source

American Political Science Review Dataverse

Collection Method

Experimental data from randomized assignments in online studies using LLM-generated counterarguments.

Freshness

Last updated 2026-06-18 19:21:31; freshness should be verified