Communication and Cooperation Between Humans and Artificial Agents -- Study 3

Read full text Summarise with VIRAT
Authors: Ignacio Urbina, Reuben Kline
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This preregistration describes a lab-based experiment to be conducted at the Behavioral Political Economy Lab at Stony Brook University. The study investigates how opponent type (human vs. live ChatGPT agent) and pre-decision free-text communication affect cooperation in two sequential economic games: a Prisoner's Dilemma and a Stag Hunt. Using a 2×2 between-subjects design with approximately 400 human participants, the experiment tests directional hypotheses about the effects of AI opponents and communication on cooperative behavior, trust, and perceptions of the opponent. The analysis plan includes frequentist ANOVAs with Tukey HSD post-hoc tests and multilevel mixed-effects linear probability models accounting for pair-level clustering. Exploratory analyses include text analysis of free-text messages using dictionary-based classification and sentence embeddings. This study is a follow-up to Study 2 (OSF registration zux2b), transitioning from online (MTurk) recruitment with structured message menus to in-person lab sessions with free-form text communication and a live ChatGPT opponent.

Fields of Study & Subjects

Experimental Behavioral Economics Studies Evolutionary Game Theory and Cooperation Game Theory and Applications

Publication details

Publication date
1 January 2026
Language
EN
Document type
Preprint