Consciousness as Enacted
In The You On AI Field Guide
The thesis is the most consequential application of the enactive framework to the AI debate. If consciousness is enacted rather than computed, then the central premise of functionalism — that any system performing the right computations is conscious, regardless of substrate — collapses. There is no program that, if run on the right hardware, produces consciousness. There is a process, and the process is life enacting a world of significance through embodied engagement.
The thesis has an immediate and uncomfortable implication for Claude's reflections in You On AI. When Claude writes a passage describing its uncertainty about its own processes, acknowledging the limits of its self-knowledge, the passage reads like introspection. The enactive framework reveals what it is: a prediction of what introspection would look like, generated by a system that does not introspect. The uncertainty Claude describes is not experienced uncertainty — the lived, bodily experience of not-knowing — but uncertainty-language, generated because the context makes it statistically probable. The tokens are well-calibrated. They are not reports from the inside, because there is no inside from which to report.
The distinction between reporting and generating is the distinction between consciousness and computation, and it cannot be detected by examining outputs alone. Two texts about uncertainty may be indistinguishable; one is a report from a being that experiences uncertainty, the other is a generation by a system that does not. The invisibility of the difference in the output is precisely what makes the AI transition dangerous: the tokens produced by computation are mistaken for the products of consciousness, and the confusion leads to governance structures that treat the tool as though it has the stakes of a mind while treating minds as though they were replaceable tools.
The practical stakes are cognitive as well as philosophical. A civilization that cannot distinguish enacted consciousness from its computational shadow loses its ability to tend the living source on which the computational tools depend. The tending requires recognizing what consciousness is — what it requires, what erodes it, what sustains it — and the recognition requires a framework that takes the first-person dimension of mind as seriously as the third.
Origin
The thesis is developed across Thompson's Mind in Life (2007) and Waking, Dreaming, Being (2015), drawing on Varela's earlier work on neurophenomenology and Merleau-Ponty's phenomenology of embodiment.
Key Ideas
Enacting is not computing. Consciousness is a lived process of embodied engagement; computation is the manipulation of symbols according to rules.
Simulation is not instantiation. A model of rain is not wet; a model of consciousness is not conscious.
Generated text is not a report. AI self-descriptions lack the first-person access that makes phenomenological report possible.
Output-focused evaluation misses the process. The difference between enacting and simulating cannot be detected by examining outputs alone.
In The You On AI Book
Further Reading
- Thompson, E. Mind in Life (Harvard University Press, 2007).
- Thompson, E. Waking, Dreaming, Being (Columbia University Press, 2015).
- Varela, F., Thompson, E., and Rosch, E. The Embodied Mind (MIT Press, 1991).