The Thermodynamic U: Refusal-Driven Dimensionality Reduction from Cosmos to Hyper-Hybrid Society

This preprint presents a unified thermodynamic framework for the evolution of the universe and humanity's place within it, extending Refusal-Driven Dimensionality Reduction Theory (RDRT). According to RDRT, phenomenal consciousness, selfhood, free will, and symbolic structures emerge as obligatory residues from enforced halts ("refusals") in recursive self-prediction processes, driven by finite energetic resources in bounded systems. Human mind represents the absolute minimum of a key parameter—subjective mental temperature (T_M)—achieved during the descent phase: from the primordial refusal in a pre-geometric vacuum (the Big Bang) through quantum-physical and biological transitions to individual finitude, with T_M in the range 0.5–2.0. In the ascent phase, however, this parameter increases: through linguistic and social structures (oscillations of T_M in collective systems) toward hyper-hybrid societies integrating artificial intelligence, where recursion depth and T_M may theoretically return to initial cosmic-scale values (relic equivalents adapted to mental scales). This suggests a cyclic dynamic in which entropy and recursion expand to universal scales (10^{30} bits and n_ref ~100–300). Time emerges as a dual residue of refusal: the objective arrow arises from global irreversible dissipation, while subjective flow is indexed by Ψ ≈ (T_M × π_context) / M_safety; both aspects stem from the same thermodynamic mechanism, providing a unified account bridging cosmic and experiential temporality. The framework is grounded in a negative ontology, in which existence and phenomenal realities are not derived from positive primitives (matter, information, or mind) but from generative absences—the thermodynamically enforced "no" of refusal, which is ontologically prior and gives rise to all structured being through its residual scars. The work integrates established data from neural energetics (Attwell & Laughlin 2001), predictive processing (Iglesias et al. 2013), cosmology (Bousso 2002), and cultural studies (World Values Survey; Hofstede dimensions). It offers falsifiable predictions, including correlations >0.7 between T_M proxies and cultural innovation indices, together with implications for consciousness research, social policy, and AI ethics in an era of accelerating hybridisation.