Creating empirically based theories of consciousness faces two obstacles: the search for neural correlates lacks working hypotheses about their causal connection to conscious states, and there is insufficient evidence that all conscious phenomena are ontologically unified. These issues may stem from the sciences lacking an engineering-level analog, like the radio engineering level that bridges theoretical electrodynamics and device function. A computational approach, describing the subject as computational primitives generating conscious states, could fill this role, as non-computational theories remain metaphysical. The paper reviews a theory based on active inference, suggesting that a computational model underlying a good theory of consciousness should be probabilistic, not deterministic.
Empirical theories of consciousness remain in a pre-paradigmatic stage, so they do not yet pose an existential threat to the philosophy of consciousness, though they demand its attention. The author clarifies the ambiguous term 'consciousness', separating aspects already open to science and technology from those unlikely to be explained away. The relationship between philosophy and science is analyzed through their inner dynamics of theories and ontologies, showing that the distinction between them is more important for science. Philosophical schemas claiming to be 'experiential' must meet criteria for empirical theories to explain anything. The author adds a pragmatic criterion: a winning theory must enable production and control of artificial conscious devices.
Consciousness is examined as a mechanism that helps living systems maintain a stable, non-equilibrium state by continuously generating and updating predictions based on sensory input. Organisms operate as probabilistic models that minimize prediction errors and optimize interaction with the environment. The article discusses Bayesian cognitive science, applying principles of Bayesian inference to understand cognitive processes, and shows how predictive processing and free energy minimization contribute to innovative theories of consciousness. It provides a brief overview of existing theories that derive consciousness from hierarchical Bayesian inference, and outlines aspects for a future unified theory, emphasizing an interdisciplinary approach.