Social imitation is often described with the language of energy. One person adopts an opinion because it appears socially advantageous, a behavior spreads because it is reinforced by its surroundings, or a group settles into a shared convention as though it were relaxing toward a preferred state. A new study by Irisarri, Trigal, Toral and colleagues challenges the limits of that analogy, arguing that the physics of social imitation cannot be fully understood by borrowing energetic concepts alone. Published in Nature Communications, the work develops a stochastic-thermodynamic perspective for social systems in which uncertainty, irreversibility and information exchange play central roles.
The research addresses a deceptively simple question: what does it mean to measure disorder, dissipation or irreversibility in a population whose members imitate one another? In conventional thermodynamics, systems exchange energy with their environment and tend toward equilibrium under well-defined physical laws. Social systems, by contrast, are driven by observations, beliefs, communication, memory and changing incentives. People do not merely transfer energy when they influence each other. They exchange information, respond to fluctuating contexts and may alter the very rules that govern future interactions.
Stochastic thermodynamics provides a mathematical language for such fluctuating systems. Rather than describing only average behavior, it follows individual trajectories through time and assigns probabilities to different sequences of events. A person changing an opinion, an agent switching strategies or a community moving between collective states can be treated as a random process. The framework then compares the probability of a trajectory occurring forward in time with the probability of its time-reversed counterpart. When these probabilities differ, the process displays a form of irreversibility that can be quantified even when no conventional energy landscape exists.
That distinction is crucial for models of imitation. Many established approaches represent social influence through an effective energy or potential: configurations in which many agents agree are assigned lower energy, while disagreement is treated as costly. Such models can reproduce striking collective phenomena, including sudden shifts between consensus and polarization. Yet the energy metaphor may conceal how those states are reached. Two systems can have similar distributions of opinions while differing profoundly in their dynamics, information flows and responses to external changes. The new study places those dynamical features at the center of the analysis.
The authors’ approach extends stochastic thermodynamics beyond systems governed exclusively by energetic exchanges. In this broader view, entropy production is not interpreted simply as heat released into a physical environment. It can also represent the statistical cost of maintaining an irreversible pattern of social transitions. If an opinion change is much more likely in one direction than the reverse, the asymmetry carries information about the mechanisms driving the process. External signals, directed influence, network structure and nonreciprocal interactions can all contribute to this irreversibility.
The framework may be especially important for understanding social systems that operate far from equilibrium. Online platforms, for example, are continuously supplied with new information, recommendations and emotionally charged content. Users influence one another, but they are also exposed to algorithms, institutions and events outside the network. Under these conditions, a population may never settle into a stable equilibrium. Instead, it can exhibit persistent probability currents, with collective states continually circulating through the system. These currents are invisible in a static snapshot but become apparent when the sequence of transitions is examined.
A technical advantage of the stochastic-thermodynamic formulation is that it separates several ideas that are frequently conflated. A system can possess a highly ordered collective state without being close to equilibrium. It can show low variability in its visible behavior while consuming information or remaining sensitive to hidden variables. Conversely, a population may fluctuate widely without producing substantial irreversibility if its transitions remain statistically balanced. By tracking path probabilities, researchers can distinguish order from equilibrium and noise from genuine directional organization.
The study also highlights why social imitation cannot be reduced to a simple tendency toward agreement. Imitation may amplify a signal, but it can also create delays, feedback loops and instabilities. When agents respond to one another at different rates, or when influence is asymmetric, the resulting dynamics may generate cycles rather than convergence. A minority can sometimes shape the trajectory of a larger group if its information arrives earlier, spreads through more influential channels or is repeatedly reinforced. In this setting, the relevant question is not only which state is most probable, but how probability moves between states and what sustains that movement.
By framing social dynamics in terms of fluctuations, information and irreversibility, the work offers a route toward more precise comparisons between physical and social systems without treating people as particles. The analogy to thermodynamics becomes a flexible set of tools rather than a claim that opinions possess literal energy. Such tools could help researchers analyze collective decision-making, cultural transmission, financial imitation and the spread of behaviors in digital networks. They may also provide new ways to identify when a system is being driven by external information, when feedback is creating instability and when apparent consensus masks an underlying flow of influence.
The broader message is that the science of imitation is entering a more dynamic phase. Social order is not necessarily the result of a population finding its lowest-energy configuration; it may be continuously produced by information exchange, asymmetric influence and fluctuations. The study by Irisarri, Trigal, Toral and colleagues places those processes within a unified mathematical framework, suggesting that the next generation of models will need to measure not only what people believe, but also the direction, probability and thermodynamic cost of how beliefs change.
Subject of Research: Stochastic thermodynamics of social imitation and collective social dynamics beyond energetic models
Article Title: Stochastic thermodynamics of social imitation beyond energetics
Article References: Irisarri, L., Trigal, L., Toral, R. et al. “Stochastic thermodynamics of social imitation beyond energetics.” Nature Communications (2026). https://doi.org/10.1038/s41467-026-76212-0
Image Credits: AI Generated
DOI: 10.1038/s41467-026-76212-0
Keywords: stochastic thermodynamics, social imitation, collective behavior, entropy production, irreversibility, information flow, nonequilibrium systems, opinion dynamics, statistical physics
