Social physics in the age of artificial intelligence
Organizations: School of Computing, Engineering and Digital Technologies, Teesside University, Middlesbrough, UK. · Google DeepMind, London, UK · Machine Learning Group, Universit´e Libre de Bruxelles, Brussels, Belgium · AI Lab, Vrije Universiteit Brussel, Brussels, Belgium · Center for Human-Compatible AI, UC Berkeley, USA · Max Planck Institute for Human Development, Center for Humans & Machines, Berlin, Germany · Informatics Institute, University of Amsterdam, Amsterdam, The Netherlands · Faculty of Natural Sciences and Mathematics, University of Maribor, Maribor, Slovenia · Community Healthcare Center Dr. Adolf Drolc Maribor, Maribor, Slovenia · Department of Physics, Kyung Hee University, Seoul, Republic of Korea · University College, Korea University, Seoul, Republic of Korea · University of Milan Biccoca, Milan, Italy
Abstract
Artificial intelligence (AI) systems are rapidly becoming more capable, autonomous, and deeply embedded in social life. As humans increasingly interact, cooperate, and compete with AI, we move from purely human societies to hybrid human-AI societies whose collective dynamics cannot be captured by existing behavioural models alone. Drawing on evolutionary game theory, cultural evolution, and Large Language Models (LLMs) powered simulations, we argue that these developments open a new research agenda for social physics centred on the co-evolution of humans and machines. We outline six key research directions. First, modelling the evolutionary dynamics of social behaviours (e.g. cooperation, fairness, trust) in hybrid human-AI populations. Second, understanding machine culture: how AI systems generate, mediate, and select cultural traits. Third, analysing the co-evolution of language and behaviour when LLMs frame and participate in decisions. Fourth, studying the evolution of AI delegation: how responsibilities and control are negotiated between humans and machines. Fifth, formalising and comparing the distinct epistemic pipelines that generate human and AI behaviour. Sixth, modelling the co-evolution of AI development and regulation in a strategic ecosystem of firms, users, and institutions. Together, these directions define a programme for using social physics to anticipate and steer the societal impact of advanced AI.
Figures & tables
| Research Area | Specific Question | Potential Design |
|---|---|---|
| Evolutionary dynamics of social behaviours in hybrid human-AI systems | (1) How can we model humans’ expectations about AI behaviour in heterogeneous, legal, normative, and hierarchical hybrid systems? (2) How can we connect short-term behavioural experiments with long-term evolutionary dynamics in hybrid populations? | Evolutionary game and statistical-physics models (e.g. replicator dynamics, stochastic processes) calibrated with experiment and field data; hybrid human–AI experiments varying AI fraction, transparency, and coordination between AI agents. |
| Machine culture: machine-generated and mediated cultural dynamics | (1) How does algorithmic curation of content and social ties affect cultural diversity, norm formation, and exploration–exploitation trade-offs? (2) How do human preferences and platform incentives shape the evolution of machine behaviour and AI-mediated culture? | Agent-based and network models of cultural transmission under different recommender objectives; A/B tests on engagement vs diversity; simulations with generative models as cultural innovators; analysis of human-in-the-loop training (e.g. RLHF) as cultural selection on AI. |
| Co-evolution of language and behaviour | (1) How does linguistic framing, when mediated by LLMs, shape human and AI decisions in strategic interactions? (2) How do language and social behaviour co-evolve in hybrid populations? | Language-based game-theoretic models; large-scale framing experiments; societies of LLM powered simulations; meta-analyses linking linguistic features to behaviour. |
| Evolutionary dynamics of AI delegation | (1) Does delegation to AI agents increase or decrease propensity towards cooperation, honesty, etc.? (2) How do incentives and AI agent design impact long-term dynamics of cooperation, inequality, honesty norms, etc.? | Simulations using simple versus LLM-based agents; Lab-based and online experiments involving humans and AI agents in dynamic, repeated interaction. |
| Evolution of epistemic pipelines in hybrid human–AI societies | (1) Under what conditions does selection favour agents with different epistemic pipelines (e.g. grounded vs text-only, causal vs correlational, cautious vs overconfident)? (2) When do hybrid systems converge to epistemic vigilance vs equilibria dominated by persuasive but ungrounded outputs? | Extended evolutionary models where types encode cognitive/epistemic parameters (grounding, counterfactual reasoning, confidence, abstention); simulations of hybrid human–AI populations; experiments varying incentives for verification vs persuasion and measuring downstream trust and misinformation. |
| Co-evolutionary dynamics of AI development and regulation | (1) How do the strategic interactions between firms, users, and regulators shape the trajectory of AI capabilities, safety investments, and power concentration? (2) Which regulatory and institutional interventions can steer this ecosystem toward safe and socially beneficial outcomes? | Game theory and population dynamics models of AI labs, regulators, and users (arms races vs cooperation; open-source vs proprietary); scenario analysis of policy levers (standards, audits, windfall-sharing); experiments and LLM-based simulations of AI actors’ decisions. |