Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory
2025/07/03 by Kenneth Payne, Payne, Kenneth, Baptiste Alloui-Cros +1 · 15 voices · 3 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #Computer Science and Game Theory (cs.GT) #FOS: Computer and information sciences #cs.AI #cs.CL #cs.GT
paper · pdf · doi:10.48550/arxiv.2507.02618
Abstract
Are Large Language Models (LLMs) a new form of strategic intelligence, able to reason about goals in competitive settings? We present compelling supporting evidence. The Iterated Prisoner's Dilemma (IPD) has long served as a model for studying decision-making. We conduct the first ever series of evolutionary IPD tournaments, pitting canonical strategies (e.g., Tit-for-Tat, Grim Trigger) against agents from the leading frontier AI companies OpenAI, Google, and Anthropic. By varying the termination probability in each tournament (the "shadow of the future"), we introduce complexity and chance, confounding memorisation. Our results show that LLMs are highly competitive, consistently surviving and sometimes even proliferating in these complex ecosystems. Furthermore, they exhibit distinctive and persistent "strategic fingerprints": Google's Gemini models proved strategically ruthless, exploiting cooperative opponents and retaliating against defectors, while OpenAI's models remained highly cooperative, a trait that proved catastrophic in hostile environments. Anthropic's Claude emerged as the most forgiving reciprocator, showing remarkable willingness to restore cooperation even after being exploited or successfully defecting. Analysis of nearly 32,000 prose rationales provided by the models reveals that they actively reason about both the time horizon and their opponent's likely strategy, and we demonstrate that this reasoning is instrumental to their decisions. This work connects classic game theory with machine psychology, offering a rich and granular view of algorithmic decision-making under uncertainty.
Cited by
Discussions
- arxiv.org/abs/2507.02618 The paper shows that different models behave completely differently when placed in game theory settings. [bsky, 3 points, 1 comments]
- Strategic Intelligence in LLMs: Evidence from Evolutionary Game Theory [hn, 2 points, 0 comments]
- Strategic Intelligence in Large Language Models: Evidence from Evolutionary GT [hn, 2 points, 0 comments]
- Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory https://arxiv.org/abs/2507.02618?utm_source=substack [bsky, 2 points, 0 comments]
- 2/2 arxiv.org/abs/2507.02618 [bsky, 1 points, 0 comments]
- Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory arxiv.org/abs/2507.02618 [bsky, 1 points, 0 comments]
- Researchers made LLMs from OpenAI, Google and Anthropic play 140,000 games of Prisoner's Dilemma. The models developed distinctive styles, and the researchers concluded they went beyond pattern matchi [bsky, 1 points, 0 comments]
- Oh, goodie. I don't know about you, but I'm trying to get on their good side. "Our results show that LLMs are highly competitive, consistently surviving and sometimes even proliferating in these compl [bsky, 1 points, 0 comments]
- Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory https://arxiv.org/abs/2507.02618v1 #AI #agents #styles #personalities (screenshot attached here is comment by Ja [bsky, 1 points, 0 comments]
- #MLSky Direct link to the paper: arxiv.org/pdf/2507.02618 [bsky, 1 points, 0 comments]
- Strategic Intelligence in Large Language Models [hn, 1 points, 0 comments]
- And Skynet becomes ever closer to reality. arxiv.org/pdf/2507.02618 [bsky, 0 points, 0 comments]
- ¿Son iguales todos los modelos de IA? Un estudio de King's College y Oxford los enfrentó al dilema del prisionero iterado y reveló sus personalidades: 🤖 Gemini: maquiavélico 🤝 Claude: conciliador 😊 [bsky, 0 points, 0 comments]
- LLMs are doing a nice job of making academic papers both accessible through real-time tutoring, but also elevating them to primary news events. My favorite headline today, "AI Develops Strategic Intel [bsky, 0 points, 0 comments]
- Researchers made LLMs from OpenAI, Google and Anthropic play 140,000 games of Prisoner's Dilemma. The models developed distinctive styles, and the researchers concluded they went beyond pattern matchi [bsky, 0 points, 0 comments]
Related