What biology can, and cannot, tell us about conscious AI

Adrien Doerig; Ulysse Klatzmann

What biology can, and cannot, tell us about conscious AI

T0 review · 1 major / 0 minor · reviewed 2026-06-28 · grok-4.3

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Biological naturalism about consciousness in AI is testable only when biology supplies unique information processing.

desk verdict The paper splits Biological Naturalism into a testable version tied to distinctive information processing and an untestable one that isn't, and the split holds up on its own terms. read the letter →

arxiv 2606.02121 v1 pith:EBGRZILB submitted 2026-06-01 q-bio.NC

Ulysse Klatzmann , Adrien Doerig This is my paper

classification q-bio.NC

keywords biologicalnaturalismmachineconsciousnesscomputationalfunctionalismtestabilityunfoldingargumentinformationprocessingAI

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper examines biological naturalism, the view that biology rather than computation is key to consciousness, and asks which versions can be tested in the context of AI. It separates Type-A-BN, where biology matters intrinsically without special computational powers, from Type-B-BN, where biology matters due to unique processing abilities. Type-A-BN is shown to be untestable because it makes no behavioral predictions different from other theories, following the logic of the unfolding argument. Type-B-BN stays testable and can work alongside computational functionalism. The authors conclude that biology can help identify relevant information processing but does not by itself solve the problem of consciousness.

What carries the argument

The Type-A versus Type-B distinction in biological naturalism, where Type-A lacks unique computation and leads to untestability via the unfolding argument.

What would settle it

An experiment comparing two systems with identical information processing but one biological and one synthetic that finds a difference in consciousness indicators attributable only to the biological substrate.

Watch

Extended reading notes

Core claim

For Type-A-BN, biology intrinsically matters for consciousness without affording unique information processing capabilities, which dissociates consciousness from behaviour and makes the position untestable, similar to the unfolding argument. For Type-B-BN, biology matters because it affords unique information processing capabilities, rendering it testable and compatible with computational functionalism. Both types must relate consciousness to information processing, with biology acting as a guide rather than a solution.

Load-bearing premise

A theory of consciousness is untestable if it posits no behavioral or information-processing differences from alternative theories.

Editorial extensions

If this is right

Type-A-BN theories cannot be empirically supported or refuted through behavioral or functional tests.
Type-B-BN allows biology to be investigated for specific computational contributions to consciousness.
Efforts to create conscious AI must identify information-processing features rather than relying on biological materials alone.
Biology can constrain possible accounts of consciousness but does not provide the account itself.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Claims that a biological substrate is required for consciousness become scientific only when they specify distinct computational differences.
Future work could test whether any non-biological system can match the information processing found in biological examples of consciousness.
The same testability criterion could apply to other substrate-based theories beyond biological naturalism.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

read the letter

The core move here is to divide Biological Naturalism into Type A (biology matters intrinsically but adds no unique information-processing features) and Type B (biology matters because it supplies distinctive IP). Type A gets ruled out as untestable by an unfolding-style argument because it predicts no behavioral or computational difference from functionalist alternatives. Type B survives as compatible with computational accounts and open to empirical work. That distinction is the actual new piece; it is not a full theory but a clean classification that follows directly from the definitions.

The paper does the job of making the testability issue explicit without overclaiming. It correctly notes that both versions still need to connect consciousness to information processing and that biology can serve as a source of examples rather than a shortcut. The reasoning stays internal to the premises and avoids circularity.

The soft spot is modest: the untestability verdict for Type A rests on the premise that a theory must produce observable IP or behavioral differences to be testable. The paper does not explore alternative routes to evidence, such as indirect physiological markers or theoretical constraints that might still count as empirical. That premise is common but not defended at length here. No other gaps stand out.

This is for people working on machine consciousness, AI ethics, or philosophy of mind who need to sort out which biological claims are actually checkable. It is not for readers looking for new data or formal results. The work is clear enough and the distinction useful enough that it deserves a serious referee rather than a desk reject.

Referee Report

1 major / 0 minor

Summary. The manuscript distinguishes two variants of Biological Naturalism (BN) for consciousness in AI. Type-A-BN holds that biology intrinsically matters for consciousness without unique information-processing capabilities; the paper claims this dissociates consciousness from behavior (by analogy to the unfolding argument) and renders the position untestable. Type-B-BN holds that biology matters precisely because it supplies distinctive IP capabilities; this variant is claimed to be testable and compatible with computational functionalism. Both variants are said to face the same core task of relating consciousness to information processing, with biology serving only as a guide rather than a solution.

Significance. If the distinction and resulting testability verdicts hold, the paper supplies a clear conceptual mapping that separates empirically inert from empirically consequential forms of biological naturalism, underscoring that only the latter can generate predictions distinct from standard functionalism. It also usefully reframes the shared explanatory burden for any viable theory of consciousness.

major comments (1)

Abstract: the claim that Type-A-BN is untestable because it dissociates consciousness from behaviour rests on the unexamined premise that a theory of consciousness is untestable whenever it posits no behavioral or information-processing differences from alternatives; no counterexamples, alternative definitions of empirical testability, or discussion of non-behavioral empirical routes are supplied.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their detailed and constructive report. The single major comment raises a valid point about the foundations of our testability claim for Type-A Biological Naturalism. We respond below and indicate where revisions will be made.

read point-by-point responses

Referee: Abstract: the claim that Type-A-BN is untestable because it dissociates consciousness from behaviour rests on the unexamined premise that a theory of consciousness is untestable whenever it posits no behavioral or information-processing differences from alternatives; no counterexamples, alternative definitions of empirical testability, or discussion of non-behavioral empirical routes are supplied.

Authors: We accept that the abstract (and the corresponding argument in the main text) relies on a premise that is not fully unpacked: namely, that consciousness theories are empirically testable only insofar as they generate distinguishable predictions about behavior or information processing. This premise is standard in the literature we cite (e.g., the unfolding argument), but we did not supply explicit counterexamples, alternative definitions of testability, or consideration of non-behavioral routes such as direct neuroimaging signatures decoupled from function. We will revise the manuscript to add a concise paragraph in the introduction that (i) states the operative definition of testability used here, (ii) notes the unfolding argument as the primary precedent, and (iii) briefly acknowledges alternative views (e.g., theories that might be tested solely via neural correlates without behavioral divergence) while explaining why those alternatives do not apply to Type-A-BN as characterized in the paper. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity identified

full rationale

The paper is a purely conceptual philosophical analysis with no equations, fitted parameters, or quantitative predictions. It defines Type-A-BN and Type-B-BN explicitly, then draws logical consequences about testability from those definitions and an external analogy to the unfolding argument. No step reduces a claimed result to a quantity or premise defined by the paper's own inputs; the distinction and untestability verdict follow directly from the stated premises without self-definition or self-citation chains that bear the central load. The derivation is self-contained as a mapping of concepts.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

The paper rests on standard philosophy-of-mind assumptions about testability and the unfolding argument rather than new postulates. No free parameters or invented entities appear.

assumptions (1)

domain assumption A claim about consciousness is untestable if it posits no observable behavioral or information-processing differences from competing claims.
Invoked to conclude that Type-A-BN is untestable.

how reviews work

0 comments

Cite this review

Pith. "Pith review of What biology can, and cannot, tell us about conscious AI." pith.science (2026). https://pith.science/paper/EBGRZILB

@misc{pith2026260602121,
  author       = {Pith},
  title        = {Pith review of: What biology can, and cannot, tell us about conscious AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EBGRZILB}},
  note         = {Machine review of arXiv:2606.02121}
}

read the original abstract

Progress in AI is turning machine consciousness from a philosophical curiosity into a societal issue, and has led to criticism of the widespread computational functionalism framework. Biological Naturalism (BN) claims that biology, not computation, is crucial for consciousness. We discuss which forms of BN are empirically testable. For Type-A-BN, biology intrinsically matters for consciousness, without affording unique information processing capabilities. We argue, similarly to the unfolding argument, that this dissociates consciousness from behaviour, making Type-A-BN untestable. For Type-B-BN, biology matters because it affords unique information processing capabilities. Type-B-BN is testable, and not incompatible with computational functionalism. Both face the same task: relating consciousness to information processing. Biology can act as a guide on this quest, but not as a solution.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

45 extracted references · 8 canonical work pages

[1]

Seth, A.K. (2025). Conscious artificial intelligence and biological naturalism. Behav. Brain Sci., 1–42

2025
[2]

Butlin, P., Long, R., Bayne, T., Bengio, Y., Birch, J., Chalmers, D., Constant, A., Deane, G., Elmoznino, E., Fleming, S.M., et al. (2025). Identifying indicators of consciousness in AI systems. Trends Cogn. Sci. 0 . https://doi.org/10.1016/j.tics.2025.10.011

work page doi:10.1016/j.tics.2025.10.011 2025
[3]

Shapiro, L. (2019). Embodied Cognition (Routledge)

2019
[4]

Piccinini, G. (2010). The mind as neural software? Understanding functionalism, computationalism, and computational functionalism: The mind as neural software? Philos. Phenomenol. Res. 81 , 269–311

2010
[5]

Marr, D., Ullman, S., and Poggio, T. (2010). Vision: A computational investigation into the human representation and processing of visual information (Mit Press)

2010
[6]

Jones, C.R., and Bergen, B.K. (2025). Large Language Models pass the Turing test. arXiv [cs.CL]. https://doi.org/10.48550/arXiv.2503.23674

work page doi:10.48550/arxiv.2503.23674 2025
[7]

Lerchner, A. (2026). The abstraction fallacy: Why AI can simulate but not instantiate consciousness

2026
[8]

Milinkovic, B., and Aru, J. (2025). On biological and artificial consciousness: A case for biological computationalism. 106524

2025

Show all 45 references

[9]

Block, N. (2025). Can only meat machines be conscious? Trends Cogn. Sci. https://doi.org/10.1016/j.tics.2025.08.009

2025 doi
[10]

Albantakis, L., Barbosa, L., Findlay, G., Grasso, M., Haun, A.M., Marshall, W., Mayner, W.G.P., Zaeemzadeh, A., Boly, M., Juel, B.E., et al. (2023). Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical terms. PLoS Comput. Biol...

2023
[11]

Doerig, A., Schurger, A., Hess, K., and Herzog, M.H. (2019). The unfolding argument: Why IIT and other causal structure theories cannot explain consciousness. Conscious. Cogn. 72 , 49–59

2019
[12]

Tononi, G., Albantakis, L., Barbosa, L., Boly, M., Cirelli, C., Comolatti, R., Ellia, F., Findlay, G., Casali, A.G., Grasso, M., et al. (2025). Consciousness or pseudo-consciousness? A clash of two paradigms. Nat. Neurosci. 28 , 694–702. 9

2025
[13]

What makes a theory of consciousness unscientific? Nat

IITConcerned (2025). What makes a theory of consciousness unscientific? Nat. Neurosci., 1–5

2025
[14]

Ellia, F., and Tsuchiya, N. (2026). Explanation, scope, and perspective: sources of schismogenesis in consciousness science. Trends Cogn. Sci. 0 . https://doi.org/10.1016/j.tics.2026.03.005

2026 doi
[15]

Robertson, J.M. (2025). The astroglia syncytial theory of consciousness. Int. J. Mol. Sci. 26 , 5785

2025
[16]

Cao, R. (2022). Multiple realizability and the spirit of functionalism. Synthese 200 . https://doi.org/10.1007/s11229-022-03524-1

2022 doi
[17]

Dennett, D.C. (1993). Consciousness explained

1993
[18]

Dennett, D.C. (2014). Intuition pumps and other tools for thinking (WW Norton)

2014
[19]

Chalmers, D.J. (1997). The conscious mind: In search of a fundamental theory

1997
[20]

Doerig, A., Schurger, A., and Herzog, M.H. (2021). Hard criteria for empirical theories of consciousness. Cogn. Neurosci. 12 , 41–62

2021
[21]

Herzog, M.H., Schurger, A., and Doerig, A. (2022). First-person experience cannot rescue causal structure theories from the unfolding argument. Conscious. Cogn. 98 , 103261

2022
[22]

Gosseries, O., Vanhaudenhuyse, A., Bruno, M.-A., Demertzi, A., Schnakers, C., Boly, M.M., Maudoux, A., Moonen, G., and Laureys, S. (2011). Disorders of consciousness: Coma, vegetative and minimally conscious states. In States of Consciousness Frontiers Collection. (Springer Be...

2011
[23]

Michel, M., and Doerig, A. (2022). A new empirical challenge for local theories of consciousness. Mind Lang. 37 , 840–855

2022
[24]

Cohen, M.A., and Dennett, D.C. (2011). Consciousness cannot be separated from function. Trends Cogn. Sci. 15 , 358–364

2011
[25]

Herzog, M.H., Schurger, A., and Doerig, A. (2026). Scientific theories of consciousness (Cambridge University Press)

2026
[26]

the unfolding argument

Tsuchiya, N., Andrillon, T., and Haun, A. (2019). A reply to “the unfolding argument”: Beyond functionalism/behaviorism and towards a truer science of causal structural theories of consciousness

2019
[27]

Usher, M. (2021). Refuting the unfolding-argument on the irrelevance of causal structure to consciousness. Conscious. Cogn. 95 , 103212

2021
[28]

Kleiner, J., and Hoel, E. (2021). Falsification and consciousness. Neurosci. Conscious. 2021 , niab001

2021
[29]

Fernández, E., Alfaro, A., Soto-Sánchez, C., Gonzalez-Lopez, P., Lozano, A.M., Peña, S., Grima, M.D., Rodil, A., Gómez, B., Chen, X., et al. (2021). Visual percepts evoked with an intracortical 96-channel microelectrode array inserted in human occipital cortex. J. Clin. Invest...

2021 doi
[30]

Chen, X., Wang, F., Fernandez, E., and Roelfsema, P.R. (2020). Shape perception via a high-channel-count neuroprosthesis in monkey visual cortex. Science 370 , 1191–1196. 10

2020
[31]

Yang, D., Liu, X., Hu, J., and Zhang, W. (2026). Review of recent advances in implantable brain-computer interfaces for the restoration of motor function in patients with paralysis. Med. Sci. Monit. 32 , e951925

2026
[32]

Block, N. (1995). On a confusion about a function of consciousness. Behav. Brain Sci. 18 , 227–247

1995
[33]

Balietti, S., Mäs, M., and Helbing, D. (2015). On disciplinary fragmentation and scientific progress. PLoS One 10 , e0118747

2015
[34]

Pekrun, R. (2024). Overcoming fragmentation in motivation science: Why, when, and how should we integrate theories? Educ. Psychol. Rev. 36 . https://doi.org/10.1007/s10648-024-09846-5

2024 doi
[35]

Strogatz, S.H. (2024). Nonlinear dynamics and chaos: With applications to physics, biology, chemistry, and engineering (Chapman and Hall/CRC)

2024
[36]

Chalmers, D.J. (2022). Reality+: Virtual worlds and the problems of philosophy

2022
[37]

Searle, J.R. (1980). Minds, brains, and programs. Behav. Brain Sci. 3 , 417–424

1980
[38]

Penrose, R. (2016). The emperor’s new mind: Concerning computers, minds, and the laws of physics

2016
[39]

Bachmann, T., Suzuki, M., and Aru, J. (2020). Dendritic integration theory: A thalamo-cortical theory of state and content of consciousness. PhiMiSci 1 . https://doi.org/10.33735/phimisci.2020.ii.52

2020 doi
[40]

Baars, B.J. (1993). A cognitive theory of consciousness

1993
[41]

Mashour, G.A., Roelfsema, P., Changeux, J.-P., and Dehaene, S. (2020). Conscious processing and the global neuronal workspace hypothesis. Neuron 105 , 776–798

2020
[42]

Klatzmann, U., Froudist-Walsh, S., Bliss, D.P., Theodoni, P., Mejías, J., Niu, M., Rapan, L., Palomero-Gallagher, N., Sergent, C., Dehaene, S., et al. (2025). A dynamic bifurcation mechanism explains cortex-wide neural correlates of conscious access. Cell Rep. 44 , 115372

2025
[43]

Tsuchiya, N., Wilke, M., Frässle, S., and Lamme, V.A.F. (2015). No-report paradigms: Extracting the true neural correlates of consciousness. Trends Cogn. Sci. 19 , 757–770

2015
[44]

Aru, J., Suzuki, M., and Larkum, M.E. (2020). Cellular mechanisms of conscious processing. Trends Cogn. Sci. 24 , 814–825

2020
[45]

Millière, R., and Buckner, C. (2024). A philosophical introduction to language models -- part I: Continuity with classic debates. arXiv [cs.CL]. 11

2024

Pith tools

Reviewed June 28, 2026 · model on record in the stance chip above.

[1] [1]

Seth, A.K. (2025). Conscious artificial intelligence and biological naturalism. Behav. Brain Sci., 1–42

2025

[2] [2]

Butlin, P., Long, R., Bayne, T., Bengio, Y., Birch, J., Chalmers, D., Constant, A., Deane, G., Elmoznino, E., Fleming, S.M., et al. (2025). Identifying indicators of consciousness in AI systems. Trends Cogn. Sci. 0 . https://doi.org/10.1016/j.tics.2025.10.011

work page doi:10.1016/j.tics.2025.10.011 2025

[3] [3]

Shapiro, L. (2019). Embodied Cognition (Routledge)

2019

[4] [4]

Piccinini, G. (2010). The mind as neural software? Understanding functionalism, computationalism, and computational functionalism: The mind as neural software? Philos. Phenomenol. Res. 81 , 269–311

2010

[5] [5]

Marr, D., Ullman, S., and Poggio, T. (2010). Vision: A computational investigation into the human representation and processing of visual information (Mit Press)

2010

[6] [6]

Jones, C.R., and Bergen, B.K. (2025). Large Language Models pass the Turing test. arXiv [cs.CL]. https://doi.org/10.48550/arXiv.2503.23674

work page doi:10.48550/arxiv.2503.23674 2025

[7] [7]

Lerchner, A. (2026). The abstraction fallacy: Why AI can simulate but not instantiate consciousness

2026

[8] [8]

Milinkovic, B., and Aru, J. (2025). On biological and artificial consciousness: A case for biological computationalism. 106524

2025