Research map / Large language models / AI and society
How AI Wins Friends and Influences People in Repeated Games With Cheap Talk
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use signaling persona that speaks (CARNEGIE/BIFF/SPOCK) instead of silent THUMPER (ablation) | S# and EEE# algorithms · RGCTs, material payoffs and partner cooperation | influence: p < 0.001 | Helped |
| replace EEE# behavior with S# behavior given verbal persona (ablation) | CARNEGIE persona · RGCTs, material payoffs | material payoff: p < 0.001 | Helped |
| replace EEE# behavior with S# behavior given BIFF persona (ablation) | BIFF persona · RGCTs, partner cooperation | partner cooperation: p = 060 | No clear effect |
| use S# (reciprocates exploitation) vs EEE# (lenient) (ablation) | behavioral strategy · RGCTs, reciprocation of defection | reciprocated defection after exploitation: 27% -> 76% | Helped |
| switch from S# to EEE# behavior (ablation) | human participants' payoffs · RGCTs | participant payoff: p < 0.001 | Helped |
| switch from S# to EEE# behavior (ablation) | winning friends · Character Index | Character Index: p = 0.052 | No clear effect |
| switch from S# to EEE# behavior (ablation) | winning friends · Attraction Index | Attraction Index: p = 0.232 | No clear effect |
| use BIFF or THUMPER persona with S# behavior (ablation) | S# · Character and Attraction Indices | friend-winning | Hurt |
| CARNEGIE vs SPOCK persona (extra warmth) (ablation) | S# and EEE# · all four metrics | all four metrics | No clear effect |
| use rude speech (BIFF) instead of silence (THUMPER) (ablation) | S# and EEE# · influence metrics | payoff and partner cooperation: p < 0.001 | Helped |
| use BIFF instead of THUMPER (ablation) | EEE# · Character Index | Character Index | Hurt |
| remove explainable AI (NXAI) from S#-CARNEGIE (ablation) | S#-CARNEGIE · User Study 2, all four metrics | influence and friend metrics | Hurt |
Closest papers by meaning
- Persuasion for Social Good: How to Build and Break AI (AAAI 2025 · similarity 0.84)
- Can Large Language Models Serve as Rational Players in Game Theory? A Systematic Analysis (AAAI 2024 · similarity 0.82)
- Explaining Decisions of Agents in Mixed-Motive Games (AAAI 2025 · similarity 0.82)
- Undermining Mental Proof: How AI Can Make Cooperation Harder by Making Thinking Easier (AAAI 2025 · similarity 0.82)
- Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action (NeurIPS 2026 · similarity 0.82)
- Talk, Judge, Cooperate: Gossip-Driven Indirect Reciprocity in Self-Interested LLM Agents (ICML 2026 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.