Why self-evidencing is a framework rather than a finished theory
A scientific theory of consciousness should make claims specific enough to be tested and, in principle, falsified. Self-evidencing, as it appears in active inference accounts, is broader: it describes organisms as continually confirming their own models of the world through action and perception, and it treats consciousness as tied to inferences about the precision or predictability of self-evidencing outcomes [3]. That is a framing principle for building explanations, not a single mechanism that can be confirmed or ruled out by one experiment.
The distinction matters because frameworks guide which experiments seem worth doing, while theories generate predictions that can fail. One analysis of Integrated Information Theory makes a parallel point: even a phenomenology-first approach that explains consciousness from consciousness itself faces a data-fitting problem, because the available information may not be enough to decide whether its explanatory hypotheses are the best ones [2]. Self-evidencing faces a similar issue—it can accommodate many findings, but accommodating findings is not the same as being tested by them.
What experiments can actually test when self-evidencing is the starting point
Experiments cannot easily test self-evidencing directly, but they can test the specific models that self-evidencing inspires. In the active inference account, phenomenal consciousness is proposed to rest on subjective valuation—a deep inference about the precision of self-evidencing outcomes—and this proposal makes contact with phenomena such as sensory attenuation and deep self-models [3]. Studies of those phenomena can therefore support or challenge the account, even though they do not test the framework in general.
The same logic applies to altered states. Psychedelic-induced ego-dissolution has been used as an objection to theories that treat self-consciousness as necessary for consciousness, but an active inference analysis argues that these states instead corroborate the role of subjective valuation and highlight psychedelic research as a way to inform consciousness science [3]. That is a case where a specific prediction derived from the framework—about what should happen when deep self-models lose precision—can be examined against experience, rather than the framework being accepted or rejected wholesale.
What this implies for designing consciousness experiments
If self-evidencing is a framework, experiments should be designed to compare specific computational models built within it, not to seek a single decisive test of self-evidencing itself. One proposal treats self-evidencing as a way to clarify the consciousness-specific functional characteristics of 'mortal computation', while leaving conscious AI as an open question [4]. That is a model-level claim: it suggests experiments should probe whether particular functional profiles are present, rather than assuming the framework already settles where consciousness exists.
The framework also shapes how unusual experiences are studied. A computational account of near-death experiences and dark retreat uses active inference to explain vivid inner imagery, life review, and self-transcendent insight as lawful reorganizations of generative self-models under conditions of radically attenuated sensory input [5]. Here the experimental implication is to manipulate or observe those conditions—sensory attenuation, collapse of action relevance, loss of precision in autobiographical self-models—and see whether the predicted dynamics appear. The framework guides the design, but the test is of the specific generative dynamics, not of self-evidencing as a whole.
A related theoretical proposal treats self-consciousness as arising through a mirror-induced process and offers the rubber hand illusion as empirical support for a principle of percept drift [1]. That work illustrates the same pattern: a broad framework generates a concrete experimental paradigm, and the paradigm can support the framework's plausibility without proving the framework itself. The practical upshot is that self-evidencing is most useful when it is treated as a source of testable models, and least useful when it is treated as a theory that experiments could confirm in one stroke.
Sources used in this answer
Mirror-induced origination process of human self-consciousness—restructuring consciousness-based cognitive theory
Proposes a mirror-induced origination hypothesis for self-consciousness, with an eight-stage model and the rubber hand illusion as empirical support for a principle of percept drift.
Can the Integrated Information Theory Explain Consciousness from Consciousness Itself?
Argues that Integrated Information Theory's phenomenology-first approach implies a self-evidencing explanation and faces a 'self-evidencing problem' about whether its explanatory hypotheses are best.
Consciousness in active inference: Deep self-models, other minds, and the challenge of psychedelic-induced ego-dissolution
Argues that phenomenal consciousness is underpinned by subjective valuation—a deep inference about the precision of self-evidencing outcomes—and that psychedelic ego-dissolution corroborates rather than undermines this account.
Self-evidencing and consciousness-specific mortal computation.
Suggests a self-evidencing framing helps clarify the consciousness-specific functional characteristics of mortal computation, leaving conscious AI as an open question.
Consciousness at the Edge of Death: Dark Retreat, Near-Death Experience, and Generative Brain Dynamics
Develops a unified active inference account of near-death experiences and dark retreat as lawful reorganizations of generative self-models under radically attenuated sensory input.