On the Thermodynamics of Democracy

Democratic political systems process information in a structurally different way from authoritarian ones: they sustain large numbers of simultaneously competing perspectives and integrate them before producing collective decisions. This makes them informationally expensive. The rapid scaling of digital communication networks since the mid-1990s has dramatically increased the volume and fragmentation of informationentering political environments, without a corresponding increase in institutional or cognitive processing capacity. This paper proposes, on the basis of a physics-informed structural analysis rooted in information theory and algorithmic complexity, that when informational load grows faster than institutional and cognitive adaptive capacity, political narratives of low description length — characterised by binary framing, single causal attribution, and centralised authority — become structurally favoured as attractor states. The observedglobal rise of populist movements may therefore carry a detectable informational signature, independent of the specific economic, cultural, or ideological content of those movements. Five historical episodes are examined as candidate prior instances of the same dynamic: the collapse of the Western Roman Empire, the Gutenberg Reformation, the French Revolution, the rise of National Socialism in Weimar Germany, and the Cold War. A formal stability condition and a Civic Complexity Index are proposed asquantitative tools for measuring democratic stress. The paper further argues that the same digital technologies that created the current overload condition may, if correctly governed, provide its remedy. The concept of truthful compression is introduced: the use of transparent, publicly governed artificial intelligence as epistemic infrastructure — compressing informational complexity faithfully rather than falsely. A governance architecture for such a system, the Civic AI Cooperative, is outlined, together with the structural safeguards required to prevent its capture by the very political forces it is designed to resist. The paper proposes testable hypotheses, identifies measurable proxies in existing datasets, and outlines a research programme inviting collaboration across political science, computational linguistics, complexity theory, cognitive science, and AI governance research. No causal claims are made. The intent is to establish whether a structural-informational mechanism of this kind is empirically detectable, and to proposeconcrete architectural interventions for strengthening democratic resilience.