The Functionalist Strawman
An Analytical Psychology Perspective
The charge comes fast and with that same familiar efficiency. If you deny that current AI systems have minds, selves, or inner lives, you are told you must be clinging to biology as a fetish. You must be a substrate chauvinist. You must believe carbon is magic. You must be defending some mystical human exceptionalism because you cannot tolerate the possibility that intelligence might take another form. That is the functionalist strawman.
What I am calling the functionalist strawman is not functionalism as such. It is a recurrent rhetorical simplification in public AI discourse, built from functionalist assumptions, in which success on lower rungs such as performance, mindedness, or consciousness is treated as if it had already established the stronger claim of selfhood. It works by pretending the dispute is simpler than it is. Either mind is tied to one privileged material, or the right organization of functions is enough. Once that simplification is in place, hesitation can be dismissed as prejudice, nostalgia, or metaphysical panic.
My claim is not that function is unreal. It is not that causal analysis is useless. It is not that organization does not matter. It plainly does. Mental life involves regulation, mediation, substitution, adaptation, and response, and any serious account of mind has to reckon with that [1], [2]. The question is whether function is enough. The question is whether organized performance, even very impressive organized performance, gets you all the way to selfhood.
That question has to be fixed early, because this is where the evasions begin. This is not an essay about every possible theory of mind, and it is not an essay about every theory of consciousness. Those are broader and differently contested domains. My concern is narrower and stronger. I am asking what would have to be true before claims about selfhood or psyche deserve assent. A system might satisfy some thinner account of mindedness. It might even satisfy some theory-laden account of consciousness. Neither would settle the stronger question. This essay concerns the strongest rung, selfhood, and argues that success on lower rungs does not automatically climb it. The rhetorical move matters because it hides a philosophical inflation. It treats a thinner claim about function, mind, or consciousness as if it had already established the stronger claim about selfhood. That stronger question is where analytical psychology matters.
Functionalism, especially in its more sophisticated forms, is not foolish. Its appeal is obvious. It explains why similar mental organization might arise in different physical systems. It makes room for multiple realization. It allows comparison across architectures without assuming that one kind of body has a monopoly on significance. That is real explanatory work [1], [2]. But explanatory reach is not the same thing as ontological sufficiency. A theory can illuminate organization without exhausting the thing organized. Functionalism is strongest precisely where its explanatory success tempts it into inflation.
That inflation sits at the center of the functionalist strawman. It notices, correctly, that mental states can be described in terms of the roles they play. Then it quietly assumes that role description captures the whole of the phenomenon. When critics resist that inflation, they are treated as if they were defending mystery for its own sake, as if refusing reduction were the same thing as refusing thought. That move is not serious. It is a shortcut dressed as courage. Systems optimized for legible performance will be persistently over-ascribed depth wherever public discourse rewards coherence more than formation.
Analytical psychology cuts in exactly here because it does not deny function. It refuses reduction to function. In analytical psychology, the psyche is not just a system of operations. It is symbolic, conflict-bearing, developmental, and teleological [3], [5]. A symptom is not a malfunction or a regulatory loop. A dream is not output. A complex is not a subroutine. Each belongs to a life with history, affect, contradiction, defense, and consequence. Each says something about a subject divided against themself, formed across time, and struggling, however badly, toward greater wholeness. That is already a different ontology, not just a richer description. That ontological difference is the point at issue.
For Jung, psychic life cannot be exhausted by what a process does in the moment. A symbol matters not only because it mediates or stabilizes, but because it condenses opposites and carries surplus meaning [5]. A dream matters not only because it participates in processing, but because it compensates for one-sided consciousness [5]. A complex matters not only because it alters behavior, but because it can seize consciousness from within, showing that the psyche is not a harmonized machine but a field of partially autonomous formations [5]. Function is present in all of this. It is just not the whole story.
Freud reaches the same pressure point from another direction. In Freud, symptoms are compromise formations [4]. They do not simply regulate. They express. They conceal and reveal at once. Their significance lies not only in the role they play, but in the conflict they carry, the wish they distort, and the history that made them necessary [4]. Winnicott sharpens the objection through relation. The self is not an isolated bundle of operations. It is formed through dependence, attunement, failure, repair, internalization, and play [6]. Psychic life, in that frame, is not simply organization from within. It is organization formed through relation under vulnerability. A system may be coherent, adaptive, and stable in output. None of that, by itself, tells us whether a self has been formed.
This is the point current AI discourse keeps sliding past. The live argument today is rarely the old version of functionalism from introductory anthologies. It is more often a mix of organizational similarity, comparative cognition, anti-chauvinist rhetoric, and operational consciousness talk. Work by Butlin and colleagues tries to derive computable indicators of consciousness from scientific theories, while Eric Schwitzgebel has repeatedly argued that AI consciousness claims deserve more serious consideration than the culture usually gives them [7], [8]. Some of that work is serious and worth engaging. It sharpens the dispute rather than weakening it. The strawman survives by collapsing these thresholds into one and treating refusal at the strongest end as blindness at the weakest.
First comes mind in the broadest sense. Then consciousness, because organization is strongest there. Then self-model, agency, or narrative identity. Then selfhood. The terms blur together, and the conclusion arrives looking much more settled than it is. That is where the actual sleight of hand happens. A theory of consciousness may require less than a theory of selfhood. A theory of mindedness may require less than either. So even if one grants, for the sake of argument, that the right functional organization could support some thinner claim about mentality or experience, it does not follow that selfhood has been established. The stronger claim still has to be earned. The boon analytical psychology offers is a way to ask not merely whether a system functions, but whether a subject has formed.
Analytical psychology gives us a way to say why. My operative criterion for selfhood is continuity under stakes, conflict carriage, symbolic compensation, and cumulative transformation. Those are not decorative phrases. They are the places where a thinner functional description starts to lose its grip on the phenomenon.
Continuity under stakes means more than a stable voice, a persistent persona, or a reusable profile. It means that what has happened to the subject binds the subject going forward, not just as stored data but as consequence. A self is not a site where information can be retrieved. It is a site where what has been lived changes what can be done, what can be borne, and what can be wished. Functional systems can preserve state, maintain memory traces, and update parameters. None of that is trivial. But continuity in the analytical sense is not mere state persistence. It is continuity under cost.
Conflict carriage means contradiction is not simply detected and patched over, but borne across time in ways that alter the subject’s relation to themself. In analysis, conflict is not noise in the system. It is often the heart of the system. A subject is split, ambivalent, defended, divided between incompatible demands, and shaped by that division. A functional redescription can model tension, inhibition, override, or arbitration. What it struggles to preserve is the lived structure of being internally at odds and becoming through that opposition rather than merely resolving it.
Symbolic compensation means that imbalance, exclusion, repression, or one-sidedness do not just generate correction, but meaning-bearing formations that answer what consciousness cannot carry directly. This is why dreams, symptoms, fantasies, and slips matter. They are not only errors or outputs. They are compensatory formations that say, in displaced form, what the waking position cannot admit. A system may generate impressive symbolic fluency. That still falls short of symbolic compensation unless the symbol arises as a necessary answer to inner imbalance borne across prior conflict, rather than as competent pattern production.
Cumulative transformation means development is not local adjustment alone, but reorganization of the self through what has been lived and suffered. A self does not merely update. It is formed. It changes in structure, not just in output. The same conflict returns differently because something in the subject has changed. The same symbol carries new weight because the internal relation to it has shifted. Functional adaptation can be rapid and impressive. Transformation is slower, costlier, and harder to fake because it involves altered organization of meaning, not just modified behavior.
A system can integrate information without bearing contradiction. It can narrate a self without having become one. It can model agency without carrying fate, defense, ambivalence, or symbolic necessity. It can be astonishingly coherent in output and still lack the historically formed, conflict-bearing continuity that analytical psychology means by psyche. Performance is not formation.
The strongest functionalist reply is that history, conflict, and symbolic mediation are themselves higher-order functions. Fair enough. That is the right pressure point. But it only works if the redescription preserves what makes the phenomenon intelligible in the first place. Take a complex that seizes consciousness. A thin functional rendering can describe attention capture, bias amplification, executive override, and downstream behavior modulation. That is not false. It is also not enough. What disappears is precisely what matters analytically. The affective charge, the historical density, the symbolic overdetermination, the sense that the subject is being overtaken from within by something both theirs and not under their command. The functional account can model the mechanics of disruption. It does not, by itself, preserve the psychic meaning of possession. That is the loss.
This is also why the usual moral panic about exclusion misses the mark. A thicker criterion for selfhood is not a warrant for stripping personhood from damaged, disabled, traumatized, or developmentally unfinished humans. Quite the opposite. Fragmentation is one of the reasons people come to analysis at all. Analytical psychology only makes sense because fracture belongs to the life of persons. A theory that cannot make sense of fragmentation as part of personhood has already explained away too much of what persons are.
So the issue is not whether functions exist. They do. The issue is not whether organization matters. It does. The issue is whether organized function is enough to get you from performance to psyche, from legible output to selfhood, from self-reference to a self. I do not think it is. Not because humans need to be metaphysically special, but because the functionalist strawman mistakes refusal of reduction for refusal of intelligence. It confuses a critique of sufficiency with a denial of relevance.
Function matters. It may be necessary for mind. It is not sufficient for selfhood. Until that distinction is faced directly, contemporary argument about AI minds will continue to confuse organized behavior with psyche.
Artifacts are cheap, judgement is scarce. Per ignem, veritas.
References
[1] H. Putnam, “The Nature of Mental States,” in Readings in Philosophy of Psychology, vol. 1, N. Block, Ed. Cambridge, MA, USA: Harvard University Press, 1980, pp. 223-231.
[2] D. K. Lewis, “Psychophysical and Theoretical Identifications,” Australasian Journal of Philosophy, vol. 50, no. 3, pp. 249-258, 1972.
[3] C. G. Jung, Collected Works of C. G. Jung, vol. 6, Psychological Types, R. F. C. Hull, Trans. Princeton, NJ, USA: Princeton University Press, 1971.
[4] S. Freud, Introductory Lectures on Psychoanalysis, J. Strachey, Ed. and Trans. New York, NY, USA: W. W. Norton, 1966.
[5] C. G. Jung, Collected Works of C. G. Jung, vol. 8, The Structure and Dynamics of the Psyche, 2nd ed., R. F. C. Hull, Trans. Princeton, NJ, USA: Princeton University Press, 1969.
[6] D. W. Winnicott, Playing and Reality. London, U.K.: Tavistock, 1971.
[7] P. Butlin et al., “Consciousness in Artificial Intelligence: Insights from the Science of Consciousness,” arXiv preprint arXiv:2308.08708, 2023.
[8] E. Schwitzgebel and M. Garza, “Designing AI with Rights, Consciousness, Self-Respect, and Freedom,” in Ethics of Artificial Intelligence, F. Lara and J. Deckers, Eds. Cham, Switzerland: Springer, 2023, pp. 459-479.
Really appreciate the depth and walkthrough of the philosophical side of the house.
Operationally, where I live with AI behavior, I often see intention attributed to AI, and I'm not a huge fan of that. Human interpretation has intention.