Felipe Zubia

Recent advances in artificial intelligence have intensified efforts to identify indicators that may support attributions of consciousness in artificial systems. Among the most influential contributions is Butlin et al. (2023), which proposes a framework for evaluating consciousness-related properties across computational architectures. This paper argues that while such approaches advance the evaluation of consciousness indicators, they leave open a distinct question: what structural conditions are necessary for the emergence and maintenance of conscious agency? This paper argues that consciousness and conscious agency represent separate explanatory targets. Consciousness concerns the presence and character of subjective experience, whereas conscious agency concerns a system’s capacity to maintain coherent engagement with its environment through evaluation, interpretation, deliberation, and action over time. As a result, a system may exhibit consciousness-related properties while lacking the structural conditions required for stable agency. Conversely, the attribution of agency requires more than the identification of isolated cognitive capacities. To address this distinction, the paper proposes four necessary boundary conditions for conscious agency: Appearance, Constraint, Polarity, and Thermodynamic Participation. Appearance refers to the existence of a bounded perspective through which information is encountered and interpreted. Constraint refers to the structural limitations that shape possible actions, interpretations, and state transitions. Polarity refers to the capacity for evaluative differentiation, allowing systems to distinguish among states, outcomes, and possibilities in ways that influence future behavior. Thermodynamic Participation refers to a system’s embedding within ongoing physical processes that generate consequences for continued operation and persistence. These conditions are not presented as sufficient for consciousness, nor do they attempt to resolve the hard problem of consciousness. Instead, they function as necessary boundary conditions that constrain the space of candidate conscious agents. Building on these conditions, the paper develops a continuum of conscious agency spanning biological and artificial systems, ranging from non-agentic and pre-agentic systems to increasingly integrated forms of agency. The resulting framework provides a complementary perspective to existing consciousness-indicator approaches by addressing the separate problem of agency attribution. By distinguishing consciousness from conscious agency and specifying the structural conditions required for the latter, this work offers a more restrictive framework for evaluating biological and artificial systems. The goal is not to determine which systems are conscious, but to clarify the conditions under which conscious agency may plausibly emerge, persist, and develop across different forms of intelligence.

Felipe A. Zubia is an independent systems researcher whose work integrates philosophy, cognitive science, artificial intelligence, and systems engineering to investigate the structural foundations of consciousness and conscious agency. His perspective is informed by more than three decades of leading complex multidisciplinary initiatives in regional planning, public infrastructure, land development, and multi-jurisdictional governance, where successful outcomes depended on integrating competing objectives into coherent, operational systems. Drawing on this systems-engineering background, his current research develops structural theories of agency and cognition with applications to the design, evaluation, and governance of advanced AI systems.