NOTE-187: On the Evolution, Science, and Metaphysics of Consciousness

Contribution

This is an author’s reply in a book symposium. Relative to the book, what is new is an explicit metaphysics of valence: a “hybrid materialist-functionalist identity” view (pp. 8–10), with the phenomenal and functional sides of valence identical rather than the phenomenal side adding causal power. The piece also states Veit’s meta-philosophy of concepts: all concepts are revisable, so “qualia” should be re-engineered rather than abandoned (p. 8). It separates Veit’s naturalism from both autopoietic enactivism and biopsychism. And it makes several positional clarifications. Gastropods may be non-conscious, and the multi-dimensional profiles are used “exclusively to map out conscious variations” (p. 11). Unity comes in degrees, and conscious selves may be “temporarily created ‘user illusions’” (p. 15). Consciousness is an adaptation, even if it is not reducible to a single function.

Key insight

For Veit, “feel” in hedonic feel names a functional architecture: “To ‘feel bad’ is not to host an entirely private sensation of badness but to occupy a coordinated regulatory state that ties a multicellular organism with high degrees of freedom together for goal-directed behaviour” (p. 10). The dispute with Frankish is therefore about concepts, not ontology. Both deny a private bureau de change. They differ on whether to keep the Cartesian vocabulary once it has been re-engineered, and Veit says keep it, since conceding the words concedes the territory (p. 7).

Assumptions

These are premises, not formal assumptions.

• Darwinian methodological naturalism. Consciousness should be studied like any other biological trait, using selection, functional analysis and phylogenetic diversity (p. 3). Human data are admissible evidence within a bottom-up programme.
• Gradualism. Consciousness is multidimensional and graded, not all-or-nothing.
• The pathological complexity thesis (PCT). Consciousness originated around the Cambrian as hedonic valence. Valence is a common currency for trade-offs in action selection among animals with high degrees of freedom. Other dimensions (sensory, self, temporal, unity) were added later.
• Revisable concepts. “In my naturalist meta-philosophical view all concepts are open to revision” (p. 8). Definitions that build in unrevisability are “chmess” (Dennett 2006).
• Metaphysical neutrality of the PCT. The PCT “is in principle compatible with a large variety of views on the relationship between matter and mind … dualism, panpsychism” (p. 9). Veit’s own preference is physicalist.
• Selection logic (p. 12). “If consciousness has no adaptive benefit for living systems, then it would have been selected against.” This is stated as a general premise; it holds only for costly traits, a qualification Veit adds later (“its costly neuroarchitecture wasn’t selected against”, p. 13).

Key results

These are the positions argued, since this is a philosophical reply.

• To de Weerd (pp. 1–5). De Weerd’s characterisation of the bottom-up approach, as one that must be independent of human evidence, is a “straw-man” (p. 4). A Darwinian account of eyes or immune systems still uses human data. Evolutionary considerations add constraints; they do not replace top-down paradigms. Top-down approaches, including iterative natural kind reasoning (Bayne et al. 2024; Birch 2020), stay anchored to humans and “risk mistaking what is unique to us for what is basic to consciousness itself”. Human evidence becomes “increasingly unreliable” the further a species is from us on the tree. The non-conscious-valence objection is deferred to a separate Philosophia symposium (Veit 2025c).
• To Sachs (pp. 5–6). Veit accepts Sachs’s split between autopoiesis 1.0 (Maturana and Varela) and 2.0 (Varela–Jonas–Thompson). He grants that 1.0’s view that all life is cognitive does not conflict in principle with the PCT, but objects when it is stretched to bacterial sentience (Reber). He rejects 1.0’s anti-Darwinian claims, and he rejects 2.0’s biopsychism as phenomenology-first armchair theory. He then claims more than Godfrey-Smith’s minimal “philosophy of nature”: “a philosophy of nature in the ultimate sense of understanding the world” (p. 6).
• To Frankish (pp. 6–10). (a) Agreement: the Cartesian paradigm is damaging, and Dennett’s hard question matters more than the hard problem. Veit has “defended illusionism, and even described myself as an illusionist elsewhere” but avoids the label (p. 7; Veit & Browning 2023a). (b) On terminology: Veit disagrees. Terms like “qualia”, unlike “phlogiston”, are too tied to experience to give up. Meanings evolve, as “energy”, “life”, “design” and species concepts did. Cartesians should not be allowed “definitional rights” (pp. 7–9). (c) On the bureau de change: phenomenality does not exert “causal power in addition to their function. They are identical” (p. 10). Valence is a cybernetic common currency. “Feeling is thus not a mysterious inner addition but just a description of what it is like to be such a system ‘from the inside’” (p. 10). (d) On note 4: the relata of the identity are “the origins of qualia” and “the origins of (complex) valence systems” (p. 9). Veit keeps identity theory in play alongside functionalism, pointing to Frankish’s own “consciousness is what our best empirical theory … says it is”. He calls the metaphysical question “premature” (p. 9). (e) Explanatory gaps remain everywhere in science (“the origin of life, the causes of the last financial crisis”) “but they are not mysteries” (p. 10). (f) Retort: “it is in fact Frankish that is caught in Cartesian gravity … playing the game by Cartesian rules” (p. 10).
• To Suzuki on evaluation versus prediction (pp. 10–12). Evaluation need not be of consciously experienced sensory states. Gastropods (Aplysia) are “a potential case” only, and would need high pathological complexity and motivational trade-offs before consciousness is confidently assigned (p. 11). Prediction-first is too narrow, because warmth, tiredness and aches are not exhausted by forecasting. It is also too broad, because immune systems, bacterial chemotaxis and plants predict. The same criticism is applied to Ginsburg and Jablonka’s Unlimited Associative Learning: it explains features of later consciousness, not its origin.
• To Suzuki on adaptation (pp. 12–13). Consciousness is an adaptation even if it opens new functions. Developmental constraints (DiFrisco et al.; Wagner) are “constraints on adaptationism”, not refutations. Independent evolution of consciousness-enabling ground plans in vertebrates, cephalopods and arthropods counts for adaptation. Lineages may lose consciousness, as snakes lost forelimbs.
• To Suzuki on disunity (pp. 13–15). There are two senses of disunity: multiple selves, and unity not being essential at all. Veit defends the second: unity is graded and switchable, as in the octopus’s arms, dolphins’ unihemispheric sleep and split-brain patients. He concedes Suzuki’s point that bilaterally integrated subcortical evaluative structures (the “selection triangle” SN–SC–PAG; Solms 2021) make fully separate streams unlikely, and reads this as “strong evidence here that the evaluative side of experience came first” (p. 15).
• To Chincarini (pp. 15–16). If evaluative experience is the core, welfare science is “methodologically relevant to consciousness science”. Veit reports that his current research has shifted towards animal welfare.

Claims

id	claim	strength	support
C1	An evolutionary bottom-up approach may use human data without ceasing to be bottom-up.	moderate	analogy with evolutionary biology of eyes and immune systems (pp. 3–4)
C2	Hedonic feel is identical with a cybernetic common-currency valence system, so no private bureau de change is posited.	weak	assertion plus a one-paragraph sketch (pp. 9–10); no mechanism, consumer system or evidence is given, and it defers to “mechanistic and functional neuroeconomic research”
C3	”Qualia” should be revised, not replaced.	moderate	argument from historical cases of concept change (energy, life, design, species), pp. 8–9
C4	The PCT is compatible with dualism, panpsychism and other metaphysics.	weak	assertion (p. 9). It sits in tension with C2’s identity claim and with calling dualism and panpsychism “games of chmess”
C5	Prediction-first accounts of consciousness are too narrow and too broad.	moderate	counterexamples (felt warmth and aches; predictive immune systems, chemotaxis, plants), pp. 11–12
C6	Consciousness is an adaptation, since a non-adaptive trait “would have been selected against”.	weak	informal argument (p. 12); the premise ignores neutral or low-cost by-products, a gap only partly closed by the later “costly neuroarchitecture” remark (p. 13)
C7	Bilaterally integrated subcortical evaluative systems are “strong evidence” that evaluation came first.	weak	inference from cited neuroanatomy (May et al. 2021; Mengual et al. 2016; Pérez-Fernández et al. 2017) via Suzuki; age of structure is taken as a proxy for priority of experience
C8	Unity of consciousness is graded and may switch; conscious selves may be temporary “user illusions”.	weak	intuition pumps (octopus, dolphins, split brain) and appeal to engineering efficiency (pp. 13–15); the claim that robotics has “conclusively shown” centralised control infeasible is asserted without citation
C9	Gastropods may be non-conscious despite rich evaluative capacities.	moderate	a concession that corrects a reading of the book (p. 11)

Concepts

• Pathological complexity thesis (PCT): consciousness evolved to handle “pathological complexity”, meaning the combinatorial trade-off problems of action selection in animals with many degrees of freedom. Hedonic valence is the original common currency.
• Evaluation-first: evaluative (hedonic) experience precedes sensory, self and temporal experience in evolution.
• Bottom-up evolutionary (BE) versus top-down experimental (TE) approach: de Weerd’s labels. Veit accepts that TE is ahistorical and human-anchored.
• Hybrid materialist-functionalist identity thesis: the origin of (complex) valence systems is identical with the origin of qualia. Phenomenality and function are the same thing, and the account stays open between role functionalism and neural identity.
• Chmess: Dennett’s term for internally rigorous but pointless inquiry. Veit applies it to dualism, panpsychism and idealism.
• Partial unity: integration of experience as a matter of degree that can change with the task.

Connections

The account of consciousness is evolutionary, gradualist, evaluation-first naturalism: the PCT, with a physicalist identity or functionalist metaphysics, sympathetic to illusionism but keeping phenomenal vocabulary. The criterion it offers for attributing consciousness is high pathological complexity together with motivational trade-offs that indicate a common currency (p. 11).

• LIT-103 (Frankish) is the commentary answered in pp. 6–10; read it first. On the positive view of valence the two converge. They differ on concepts (revise versus replace) and on whether identity theory stays open.
• LIT-120 (Bruckner, welfare subjects and autopoiesis) uses “autopoiesis” loosely for adaptive self-maintenance to ground welfare without sentience. Veit rejects autopoiesis as a rival origin of consciousness story but grants that autopoiesis 1.0 does not conflict in principle with the PCT. The two pieces address different questions (welfare subjecthood versus consciousness) and are compatible on that point.
• LIT-185 (IIT): Veit groups IIT with panpsychism as resting on “unquestionable (rather than revisable) axioms” that make it “no longer part of intellectual inquiry” (p. 8), though he grants it may be a useful avenue of research “under the right conditions”. This is harsher than NOTE-122’s reading, which faults IIT’s formal definability rather than its axioms’ epistemic status.
• LIT-193 (Terrill & Veit 2024, plant quasi-sentience): Veit here does not count Paramecium or Chlamydomonas as conscious (p. 11). He objects to extending cognition arguments to bacterial sentience (p. 5), calls plant ‘experiences’ predictive control that would make prediction-based criteria “implausibly liberal” (p. 12), and cites his own forthcoming “a case against plant consciousness”. That is a shift from, or at least a tension with, the co-authored 2024 openness to plant quasi-sentience. (The forthcoming paper was not read; its content is unverified beyond its title.)
• LIT-146 (Gruen) and LIT-193 share the welfare link that Chincarini and Veit endorse.
• LIT-135 (Seth, biological naturalism, skim only) ties consciousness to life via predictive processing and autopoiesis, a combination Veit rejects on both counts here.

Consciousness tag: justified. The piece is centrally about the evolution, science and metaphysics of consciousness.

Bearing on the record

• It gives the record a citable statement that the PCT’s “felt common currency” is meant functionally or cybernetically, which answers the bureau-de-change objection as a claim. Any THEORY document relying on valence-as-common-currency should cite this together with LIT-103 and note that the positive mechanism is still promissory.
• It also records Veit’s own tightening on plant and microbial consciousness, relevant wherever LIT-193 is cited.
• ML practice: none. Veit uses robotics and AI only as an aside about centralised control (p. 14). There is nothing for the Anthology of the SOTA.

Limitations

• It is a reply piece. The objections are known only through Veit’s quotations of them, and the book’s arguments are presupposed, not restated.
• The answer to Frankish’s substantive demand is a sketch. There is no account of the valence signal’s consumers, format or causal route to action.
• There is an internal tension between claiming metaphysical neutrality for the PCT (p. 9) and asserting identity of phenomenality and function (p. 10). The neutrality holds for the PCT as a teleonomic thesis; the identity is Veit’s own add-on. He does not separate the two explicitly.
• Several empirical-sounding claims are unsupported: that robotics has “conclusively shown” the infeasibility of a homunculus (p. 14), that evaluative structures show “strong evidence” of evaluation’s priority (p. 15), and that non-adaptive traits are selected against (p. 12).
• The article’s abstract miscounts the commentaries it answers (“four” for five).

Open questions

• What would a mechanistic account of valence as a common currency look like, and what evidence would show that a given animal uses one? Veit points to motivational trade-off paradigms and neuroeconomics but gives no protocol.
• Do bilaterally integrated subcortical systems predict a unified evaluative experience alongside partially disunified sensory experience? This is the testable residue of the disunity exchange.
• Where does Veit’s forthcoming “case against plant consciousness” leave the quasi-sentience position of LIT-193?

Corrections to the seeded skim

• It is not “four commentaries”. The dossier’s abstract paraphrase follows the article’s own abstract, which says “four critical replies” in both versions. The body answers five commentators: de Weerd, Sachs, Frankish, Suzuki and Chincarini. The version of record’s introduction says “five critical commentaries” (p. 1), while the preprint said “four” and then listed five names. Sachs’s commentary is in Part 1 of the special issue (Adaptive Behavior 33(3)); the other four are in Part 2. Sinha’s contribution is also discussed briefly (pp. 2, 15–16) but is not treated as one of the critical commentaries.
• Access. The dossier says no full text was reachable. The PhilSci-Archive PDF loaded this time, and the version of record is on the author’s site as “veit-2025-on-the-evolution-science-and-metaphysics-of-consciousness.pdf”.
• Preprint versus version of record. The substantive changes are these. (i) The abstract adds “and expand my evolutionary account”. (ii) “a bottom-up approach … is right after all” becomes “is the correct approach after all” (p. 2). (iii) The identity claim is narrowed: “the origins of qualia are identical with the origins of (complex) valence systems” (p. 9); “(complex)” is new. (iv) The Dennett–Goff anecdote is now attributed to “Keith Frankish’s recollection” (p. 8). (v) The funding line changes from “N/A” to “I thank Open Philanthrophy [sic] for supporting this research”. (vi) The duplicated section number “4” of the preprint is gone. The rest is copy-editing.
• The dossier’s check on Suzuki is answered, and the answer is negative. Veit distinguishes evaluation-first from prediction-first only conceptually and phylogenetically. The prediction view is “too narrow and too broad” (pp. 11–12), and the temporal dimension is “the least likely to have been present at the very origins”. He appeals to subcortical evaluative structures and Solms’s PAG (pp. 14–15). He proposes no empirical test that would tell the two apart.
• The dossier’s reading of the reply to Frankish is incomplete. The dossier says Veit answers “the charge of residual Cartesianism”. He does, but he also concedes Frankish’s central demand (“Frankish is right to demand a ‘positive naturalistic account’”, p. 9). What he supplies is one paragraph (pp. 9–10), not the account of “nature and causal powers, the systems that consume them” that Frankish asked for.
• Suggested extra tags, not a change of primary: metaphysics, since a whole section is on the metaphysics of consciousness, the identity thesis, the hard problem and panpsychism; and possibly ethics, for the welfare section. consciousness stays first.