by Stevan Harnad & ChatGPT
Preamble (Stevan Harnad):
(1) There is a fundamental difference between the “Other Minds Problem (OMP)” (that the only one who can observe that a feeler feels is the feeler) and the “Hard Problem” (HP).
(2) “Consciousness” is a weasel-word; it means too many different things to too many people.
(3) Sentience—the capacity to feel (i.e., “the capacity to be in a state that it feels like something to be in”)—is not a weasel-word. It is the only directly “observable” thing there is.
(4) By definition, all sentient entities can feel. (The obfuscating weasel-version of sentience is “phenomenal consciousness”: “felt feeling”)
(5) There is only one OMP—but there are many different kinds of feelings (from touch to taste to tremor to thinking).
(6) The capacity to feel (anything) is what the comparative evolution of “consciousness” (feeling) is really about. But this is not just the OMP (“Which species can feel? What do they feel? and How and why did that capacity evolve?”).
(7) The OMP can only be studied through its observable correlates: “What can different species detect and react to?” “What was the origin and adaptive advantage of these observable correlates of sentience?”
(8) But the evolutionary problem of explaining the origin and adaptive advantage of sentience is not just the OMP. It is also the “Hard Problem (HP)” of explaining, causally, “how and why what different species can detect and react to (i.e., what they can do) is not just detected and reacted to but felt.
Harnad & GPT:
The central difficulty in the Royal Society volume is that almost every chapter, including the editors’ framing introduction, proceeds as if the distinction drawn in these eight points did not exist. The contributors are unanimous in treating sentience as if its existence and causal potency were unproblematic: they take for granted that feeling plays adaptive roles in learning, prediction, decision and coordination. This is surely true. But the challenge is to explain how and why the neural or behavioural mechanisms they describe are felt rather than merely performed. The authors treat feeling as though it were just another biological property awaiting the same sort of explanation that feathers or kidneys receive, rather than the anomalous property singled out in point 8. Consequently, the question that the volume endeavours to address—What is the adaptive function of consciousness—is answered on an operational level only: they explain what organisms can do, not why any of these doings feel like anything.
The editors simply assume that a functional role for experience implies that experience is the cause of the function. But this merely presumes what is to be explained. It does not show why the functional capacity could not be instantiated by an unfelt mechanism, which is the substance of the Hard Problem. In their eagerness to naturalize consciousness, they treat feeling as if it were transparently part of the causal machinery, thereby glossing over the explanatory challenge they hope to confront.
The individual chapters adopt the same pattern. When Humphrey distinguishes cognitive from phenomenal consciousness through blindsight, he proposes that phenomenal experience evolved because internalized efference copies make things “matter.” But the argument only redescribes the behavioural consequences of feeling and attaches them to a proposed neural mechanism. It does not explain how efference copies become felt, nor why “mattering” cannot be functional rather than felt. The distinction between blindsight and sighted vision merely demonstrates different forms of information processing. The transition to felt vision—point 8—is treated as if it came automatically with the functional mechanism.
Similarly, Ginsburg and Jablonka propose that unlimited associative learning marks the presence of consciousness and that felt states play a role in “mental selection.” (The “mental” is somewhat redundant: why not “internal selection”?). But again, the fact that an organism learns flexibly and projects goals does not explain how if why such processes are felt. Their marker identifies a behavioural threshold, but the behavioural threshold does not itself entail or explain feeling. In linking UAL to phenomenal consciousness, they rely on the assumption that because flexible learning is sophisticated, it must be accompanied by felt experience. This conflates the OMP with the HP and leaves the causal question untouched.
Moncoucy, Tallon-Baudry and Cleeremans similarly treat phenomenal consciousness as an evolved internal valuation system. The explanatory vocabulary is motivational, computational and behavioural; feeling is assumed to be the medium of valuation, not explained. Their suggestion that pleasure becomes a proximate motivator does nothing to close the gap between reactive behaviour and felt valence. They redescribe the function of hedonic signals, but the hedonicity itself is again taken for granted.
Andrews and Miller propose that sentience evolved to support social coordination. But their argument takes for granted that the social signals in question are felt, and that without felt states the coordination would fail. This simply assumes that felt experience is necessary for the adaptive benefit, which is exactly what needs to be explained.
Crystal’s treatment of episodic memory repeats the same pattern: because episodic recollection in humans is rich, and because rats exhibit what-where-when integration, rats must likewise feel the temporality of recollection. But no causal explanation is given for how replay mechanisms become felt, nor why replay must be felt to perform its adaptive function.
Tramacere, Chittka, Kohda, Maldarelli, Güntürkün, Klein & Barron, Newen & Montemayor, Singer & Damasio all likewise attribute functions—flexibility, prediction-error integration, self-representation, action selection, analogue interoception—to states that they assume to be conscious. But in every case the feeling of those states is not explained; it is simply presupposed as part of the mechanism. When one extracts the functional core of each proposal, nothing why it need to be executed non-sentiently. This is the sense in which all these accounts beg the question of point 8.
The cumulative problem is that the authors substitute the study of behavioural and neural correlates—the only tractable approach to the OMP—for an explanation of how or why any of these correlates need be felt. They mistake empirical correlates for causal explanations. This is precisely the confusion between the OMP and HP that the eight points distinguish and that the editors do not address. The volume is therefore valuable as a survey of what organisms can do, but without a hint of why any organism feels anything in doing it. That question is presupposed at the outset and then buried under functional reinterpretations of what are, in effect, only correlates of sentience.
Fitch, W. T., Allen, C., & Roskies, A. L. (2025). The evolutionary functions of consciousness. Philosophical Transactions B, 380(1939), 20240299.