Anil Seth's Biological Naturalism and His 2026 BBS Response
Anil Seth’s biological naturalism challenges whether computation alone can be sufficient for consciousness. His 2025 Behavioral and Brain Sciences target article links consciousness to living systems. The abstract of his September 2026 response proposes research on the substrate dependence of consciousness and cognition. It does not report a concession to any named commentator.
Correction on 17 September 2026. An earlier version of this page assigned specific arguments to named commentators without verified sources. It also attributed concessions and a proposed artificial consciousness test to Seth’s response without evidence. Those passages, the commentary grouping table, and the unsupported count of 30 responses have been removed.
Anil Seth’s biological naturalism in the target article
Cambridge published the accepted manuscript of Conscious artificial intelligence and biological naturalism online on 21 April 2025. This account uses that accessible manuscript, which is not the final version. Its DOI is 10.1017/S0140525X25000032.
In section 1, Seth describes biological naturalism as the view that consciousness belongs to only, though not necessarily all, living systems. He explicitly allows that life need not be carbon based. Thus his proposal neither assigns consciousness to every organism nor rules out every possible artificial conscious system.
Sections 3.1 and 3.3 distinguish computational functionalism from substrate flexibility. The former treats suitable computations as sufficient for consciousness. The latter allows conscious states in some alternative materials, without assuming that every material can support them. Seth notes that substrate flexibility could hold even if computational functionalism were false.
Section 3.5 addresses emergence and the separation of scales. Seth discusses causal and informational closure as ways to assess whether higher level dynamics can be described independently of finer physical details. He warns against assuming that the separation available in conventional computers also holds in living systems. Section 3.6 explicitly states that the preceding arguments do not disprove computational functionalism.
The manuscript’s abstract judges consciousness unlikely along current AI trajectories. It considers consciousness more plausible as artificial systems become more like brains or living systems. These are Seth’s assessments, rather than demonstrated necessary and sufficient conditions for consciousness.
The September 2026 response abstract
Cambridge published Seth’s author response, The stuff matters, online on 17 September 2026 in volume 49, article e366. The accessible Cambridge page provides its abstract and references. We did not obtain the full response, so this section is limited to the abstract.
Seth thanks the commentators and says he will identify common themes. He stresses the increasing urgency of considering both conscious AI and the consequences of AI that appears conscious. He also identifies a possible research programme on the substrate dependence of both consciousness and cognition.
The abstract does not identify which objections Seth accepts or rejects. It does not give a test based on artificial metabolism, establish agreement among the commentators, or explain his answer to Chalmers, Koch, or Solms. A reference list cannot establish any of those claims.
Editorial analysis of the two questions
The distinction between consciousness and cognition sets two different research tasks. A substrate change could alter prediction, memory, or learning without establishing anything about experience. Conversely, a system’s apparent competence cannot settle whether its physical processes support consciousness.
The following framework is our reading of the evidential requirements. It is not a reconstruction of the unavailable response.
| Question | Evidence needed | Limit on the inference |
|---|---|---|
| Does a physical detail affect cognition? | A comparison that changes that detail and measures cognitive performance | A performance difference alone does not identify a condition for experience |
| Does that detail affect consciousness? | A justified link between the changed mechanism and evidence of consciousness | A result in one substrate does not establish necessity in every substrate |
| Does an artificial system appear conscious? | Evidence about the system’s outputs and how people interpret them | Perceived consciousness and actual experience remain distinct questions |
This separates the reservoir computing study by Tolle, Luppi, Seth, and Mediano from a test of artificial consciousness. Measuring an emergent property alongside predictive performance leaves open whether that property is sufficient for experience. It also leaves open whether the same measurement captures all relevant physical processes in different substrates.
Blackburne and colleagues’ compatibilist emergence commentary supplies a related distinction. Its abstract describes a kind of emergence as necessary but not sufficient for consciousness. Even accepting that proposal would leave a further task of identifying what else consciousness requires. Neither abstract settles that task.
What remains unresolved
Anil Seth’s biological naturalism remains a proposal to assess against alternatives in the consciousness theory index. The 2026 overview of AI consciousness research covers the wider assessment problem.
The original manuscript supplies reasons to question computational sufficiency. The response abstract names a research direction. Neither gives us grounds to announce a settled debate or a change in Seth’s position. A detailed account of his replies requires the full response and the relevant commentaries.
Sources
- Anil K. Seth. Conscious artificial intelligence and biological naturalism. Behavioral and Brain Sciences, accepted manuscript, published online 21 April 2025. Cambridge manuscript and PDF. Sections 1, 3.1, 3.3, 3.5, and 3.6 inform the account above.
- Anil K. Seth. The stuff matters (subtitle on consciousness, computation, and biology). Behavioral and Brain Sciences 49, e366, published online 17 September 2026. DOI and publisher abstract. Only the abstract and publication details support the response account.
