{
  "id": "functionalism-philosophy-of-mind",
  "title": "Functionalism (Philosophy of Mind)",
  "lang": "en",
  "source_lang": "en",
  "summary": "Functionalism identifies mental states by their causal roles rather than their material composition, linking theories of mind to computation, psychology, and debates about consciousness.",
  "category": "philosophy",
  "subcategory": "metaphysics-epistemology",
  "type": "concept",
  "aliases": [
    "Functional theory of mind",
    "Mental-state functionalism"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/functionalism-philosophy-of-mind/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/functionalism-philosophy-of-mind/",
      "json": "https://aiwiki.page/zh/functionalism-philosophy-of-mind.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Functionalism is a family of theories in philosophy of mind according to which a mental state is characterized by what it does within a system rather than by the material from which that system is made. A state such as pain, belief, or desire is identified through its relations to sensory inputs, other internal states, and behavioral outputs. Pain, for example, is typically produced by bodily injury, interacts with beliefs and desires, and contributes to avoidance, attention, and protective action. A functional account treats an appropriately specified pattern of such relations as constitutive of being in pain, not merely as evidence for an independently defined state.\n\nThe approach became prominent in the 1960s and 1970s as an alternative to both behaviorist analyses of mental concepts and theories identifying mental kinds with particular kinds of brain state. Its central attraction is that a single psychological organization might be realized in different biological structures or artificial systems. Functionalism has consequently shaped debates about physicalism, artificial intelligence, and the scientific explanation of cognition. Its principal difficulties concern whether causal organization alone can account for conscious experience, meaningful thought, and the causal efficacy of the mental."
    },
    {
      "heading": "Functional roles and mental-state identity",
      "text": "The “function” in functionalism is primarily a causal or organizational role. It is not necessarily a biological purpose, an evolutionary adaptation, or a consciously intended use. A state has a functional role because of the circumstances that tend to produce it, its interactions with other states, and the changes it helps bring about. A functional description therefore concerns an organized system of dependencies rather than a list of outward actions. Different theories disagree about how much detail this organization must include and whether its defining relations are conceptual, scientifically discovered, or partly normative.\n\nConsider thirst. A simplified account relates thirst to dehydration, awareness of an available drink, a preference for obtaining liquid, and drinking when suitable opportunities arise. Thirst does not invariably cause drinking: a person may believe the liquid is poisonous, wish to observe a fast, or be unable to move. The relevant role therefore includes interactions with beliefs, desires, perceptions, and capacities. Likewise, pain need not always produce withdrawal. A person may endure an uncomfortable procedure because of other commitments. Functional descriptions must accommodate these interactions rather than treating every departure from a stereotyped response as evidence that the mental state is absent.\n\nThis relational approach is usually holistic. Beliefs are partly characterized by their relations to desires and actions; desires are partly characterized by their relations to beliefs and actions. Such interdependence would be circular if each mental term were defined separately through unexplained appeals to the others. Functionalists instead characterize a network of states together. Just as positions in a formal game can be specified by their relations within its rules, psychological states can be identified through their places in a larger organization. The analogy concerns relational individuation, not the claim that minds literally follow explicit rules.\n\nFunctionalism must also distinguish an actual sequence of events from a dispositional organization. Two systems could produce the same behavior on one occasion while differing substantially in what they would do under other circumstances. A convincing functional equivalence therefore involves counterfactual dependencies: how the system would respond to other inputs, what would occur if an internal state changed, and which transitions depend on which components. A recording that plays an appropriate answer by coincidence does not thereby possess the organization responsible for understanding a question.\n\nThe thesis is thus stronger than the claim that mental states have useful effects. Almost every theory accepts that psychological states participate in causal patterns. Distinctive functionalism says that the appropriate role determines the mental kind itself. Whether that determination is complete—especially for experience and intentional content—is a central point of disagreement."
    },
    {
      "heading": "Historical development and intellectual background",
      "text": "Functionalism emerged from several developments within twentieth-century analytic philosophy. One was dissatisfaction with behaviorism, especially attempts to analyze mental concepts entirely through actual behavior or behavioral dispositions. Such accounts faced difficulties explaining how the same belief might lead to different actions depending on a person’s desires and other beliefs. Functionalism retained an interest in publicly investigable organization while admitting internal states as genuine explanatory intermediaries. It therefore differed both from radical restrictions on psychological explanation and from the claim that mental vocabulary merely abbreviates descriptions of behavior.\n\nA second background was the rise of the mind–brain identity theory. U. T. Place’s “Is Consciousness a Brain Process?” appeared in 1956, and J. J. C. Smart’s “Sensations and Brain Processes” in 1959. These works helped establish the possibility of identifying experiences with physical processes without treating mental and physical expressions as synonymous. Functionalists nevertheless questioned whether each psychological kind had to correspond to one neurophysiological kind. Creatures with substantially different nervous systems might plausibly share pain, memory, or perception even if their underlying physical states differed.\n\nA third influence came from mathematical computation. Alan Turing’s 1936 account of computability described machines through abstract states and rules of transition rather than through a particular engineering material. This suggested a way to distinguish an organization from its implementation. Hilary Putnam developed machine-based approaches to mentality in papers including “Minds and Machines” (1960) and “Psychological Predicates” (1967), later reprinted as “The Nature of Mental States.” These writings made the possibility of psychologically similar but physically different systems a major challenge to simple type-identity theories. Putnam’s later philosophical positions departed substantially from his early functionalism.\n\nRelated approaches developed without making a literal digital computer the universal model of mind. In A Materialist Theory of the Mind (1968), D. M. Armstrong characterized mental states through their causal relations to stimuli, behavior, and other internal states. David Lewis, particularly in “An Argument for the Identity Theory” (1966) and “Psychophysical and Theoretical Identifications” (1972), developed methods for defining theoretical terms through their roles in a larger theory. These approaches helped connect functional characterization with empirical identification of the states occupying the relevant roles.\n\nDuring the 1970s, Jerry Fodor connected functionalist ideas with arguments for the autonomy of psychology and with representational theories of cognition. His 1974 paper “Special Sciences” argued against requiring every higher-level scientific kind to receive a uniform reduction to a physical kind. Functionalism consequently became a broad research orientation rather than a single doctrine: philosophical analysis, computational psychology, and causal theories of mind shared an emphasis on organization while differing about its precise significance."
    },
    {
      "heading": "Formal models and the definition of theoretical terms",
      "text": "Machine functionalism makes the idea of a functional role precise by modeling a system as a collection of internal states connected by transition rules. A simple deterministic model can be represented as\n\n\\[M=\\langle S,I,O,\\delta,\\lambda\\rangle,\\]\n\nwhere \\(S\\) is a set of internal states, \\(I\\) a set of inputs, \\(O\\) a set of outputs, \\(\\delta:S\\times I\\rightarrow S\\) a transition function, and \\(\\lambda:S\\times I\\rightarrow O\\) an output function. The current state and input jointly determine a successor state and an output. Probabilistic models replace fixed transitions with probability distributions. The model describes an abstract organization; it does not by itself specify whether its components are neurons, electronic circuits, or something else.\n\nIn this framework, an internal state is identified by its position in the transition structure, not simply by the output associated with it on one trial. Two machines may produce identical outputs for a particular input sequence while having different internal organizations. Conversely, distinct physical mechanisms may instantiate the same transition structure. A Turing machine provides a particularly influential model, although its full computational configuration includes the tape contents and head position as well as its finite control state. Psychological modeling need not be restricted to that architecture: continuous variables, parallel processing, and feedback dynamics can also be described functionally.\n\nA second formal approach treats psychological vocabulary as theoretical vocabulary. Suppose a theory \\(T\\) contains mental-state terms \\(M_1,\\ldots,M_n\\), together with vocabulary describing inputs, outputs, and other relevant relations. Replacing the mental terms with variables and existentially quantifying over them produces a Ramsey sentence:\n\n\\[\\exists X_1\\cdots\\exists X_n\\, T(X_1,\\ldots,X_n;I,O).\\]\n\nThis states that some collection of properties or states jointly satisfies the theory. It allows the network to be characterized without first providing independent definitions of every mental term. Frank Ramsey’s method originated in the analysis of scientific theories; its application to psychology became an important tool for functionalist accounts.\n\nThe existential claim alone, however, does not uniquely define each mental state. Several different assignments might satisfy the same theory. A stronger role definition may identify a state with the property occupying a particular position in the unique appropriate realization of the theory, assuming such uniqueness exists. If no assignment satisfies the theory, or several incompatible assignments do, further interpretation is needed. Functionalists may appeal to approximation, context, a more detailed theory, or distinctions among different realizations.\n\nFormalization therefore exposes substantive choices rather than eliminating them. The theory must specify what counts as an input, which internal differences matter, and how outputs are individuated. Defining output merely as bodily movement yields a different theory from defining it as goal-directed action. Similarly, individuating inputs as retinal stimulation differs from individuating them as encounters with objects. These choices influence both the empirical adequacy and the metaphysical implications of the resulting account."
    },
    {
      "heading": "Analytic, scientific, role, and realizer functionalism",
      "text": "One major division concerns where a functional specification comes from. Analytic functionalism seeks to articulate the roles implicit in ordinary mental concepts. On this approach, understanding pain, belief, or desire already involves grasping some characteristic relations among circumstances, other mental states, and action. Philosophical analysis makes those relations explicit. The resulting definitions are intended to capture conceptual constraints rather than simply report contingent discoveries about human nervous systems. Lewis’s use of ordinary psychological theory is a central example of this general strategy.\n\nThe appeal to ordinary psychology need not mean that every commonplace opinion becomes part of a definition. People can hold mistaken beliefs about memory, perception, or motivation while retaining the relevant concepts. The challenge is to distinguish constitutive commitments from revisable assumptions. If an account makes a supposedly defining connection too strict, it may exclude familiar cases, such as pain without overt distress. If it weakens the connection excessively, it risks providing too little structure to distinguish one state from another. Analytic functionalism thus depends on judgments about which relations are essential to the concepts being analyzed.\n\nPsychofunctionalism instead draws its specifications from scientific psychology. A psychological kind may be defined by its role in the best explanatory theory, even when that role is not apparent to ordinary speakers. Models of working memory, perceptual processing, or attentional selection may posit states and mechanisms that have no direct counterpart in everyday vocabulary. On this approach, functional characterization is generally empirical and revisable. A successful theory might reorganize familiar categories rather than vindicate them unchanged. The price is that scientific roles may not answer every question about the meaning of ordinary mental terms.\n\nA different distinction separates role functionalism from realizer functionalism. Role functionalism identifies a mental property with occupying the relevant role. The property of being in pain would then be a higher-level property that different underlying states could instantiate. Realizer functionalism identifies the mental property with the property that actually performs the role in a specified system or population. Human pain might consequently be identified with a particular neurophysiological kind, while another species’ pain has a different physical realizer. Both approaches begin from functional characterization, but they locate mental properties at different levels.\n\nThese divisions cross-cut one another. An ordinary conceptual role could be used to identify an empirically discovered neural realizer, while a scientific theory could define a higher-level role shared by several mechanisms. Functionalism also does not, by itself, logically require physicalism. If a nonphysical state could enter the appropriate causal organization, an unrestricted functional definition might admit it. Most functionalists favor physical realization, but that additional commitment should be distinguished from the claim that mental kinds are functionally individuated."
    },
    {
      "heading": "Multiple realizability and the autonomy of psychology",
      "text": "The argument from multiple realizability supplies one of functionalism’s strongest motivations. Its starting point is that the same psychological capacity may occur in systems with different physical constitutions. Humans, other mammals, birds, and cephalopods differ in neural organization, yet comparisons among their perception, learning, and behavior are scientifically meaningful. Functionalists extend this possibility to hypothetical extraterrestrial organisms and artificial systems. If these systems can share a mental state without sharing one narrowly specified physical state, then identifying the mental kind with that physical kind appears too restrictive.\n\nThe argument is directed most immediately against simple type–type identities. A type is a repeatable kind, such as pain; a token is an individual occurrence, such as a particular painful episode. Rejecting one universal identity between pain and a single neural type does not entail that any actual pain token is nonphysical. A functionalist may hold that every particular episode is wholly realized by physical processes while denying that all such episodes belong to one independently identifiable neurophysiological kind. This distinction explains why functionalism has often been associated with nonreductive physicalism rather than dualism.\n\nFunctional organization also helps explain why psychological generalizations can be useful despite physical variation. A model relating perception, stored information, decision, and action may apply across systems whose detailed implementations differ. Its explanatory value depends on identifying stable causal organization, not on listing every molecule involved. Fodor’s defense of the special sciences emphasized that higher-level kinds may correspond to heterogeneous physical realizations. A disjunction of all those realizations might fail to constitute a useful kind within fundamental physics, even if every instance is physical.\n\nMultiple realizability nevertheless does not establish functionalism by itself. The inference from “not one narrow neural type” to “nothing but functional role” requires further argument. A psychological property might depend on biological, historical, or embodied features not captured by a proposed role description. Moreover, differences in neural structure do not automatically show that the same fine-grained psychological kind is present. Two species might solve superficially similar tasks through importantly different capacities. Establishing genuine multiple realization therefore requires independent criteria for sameness at the psychological level.\n\nThe possibility of local reduction further complicates the picture. A mental kind might receive a mechanistic explanation within one species, developmental stage, or neural architecture without a universal reduction across every possible thinker. Experience-dependent changes in the brain may also alter implementation while preserving a capacity only approximately. Functionalists must consequently distinguish broad metaphysical possibilities from actual comparative findings. Multiple realizability supports an organizational perspective, but the proper grain of psychological and neural classification remains a substantive scientific and philosophical question."
    },
    {
      "heading": "Intentional content and the limits of internal organization",
      "text": "Many mental states are about something: a belief concerns the weather, a desire concerns a future event, and a memory concerns an earlier experience. This directedness is called intentionality. Functionalists often explain it through the roles that representations play in inference, perception, and action. A belief that rain is approaching may interact with a desire to remain dry and a belief about the location of an umbrella to produce a particular decision. Its content is therefore connected to its place in an organized pattern of reasoning and practical response.\n\nAn immediate difficulty is that role and reference need not coincide straightforwardly. Two representations can participate in similar internal transitions while referring to different things. Conversely, people can think about the same object despite differing considerably in their associated beliefs. If every inferential connection helps constitute content, apparently minor differences in knowledge may prevent two people from sharing a belief. Such strong holism creates problems for communication, disagreement, learning, and psychological generalization. Functionalists may therefore distinguish central from peripheral connections, or distinguish a narrowly individuated cognitive role from broader semantic content.\n\nSemantic externalism presents a different challenge. Putnam’s “The Meaning of ‘Meaning’” (1975) introduced the Twin Earth example, in which an imagined liquid resembles ordinary water in familiar observable respects but has a different underlying composition. The example supports the view that reference can depend on a thinker’s environment rather than solely on internal organization. Tyler Burge’s “Individualism and the Mental” (1979) similarly emphasized the role of a linguistic community in determining thought content. If internally similar thinkers can have different contents, an exclusively internal functional specification cannot determine every aspect of their beliefs.\n\nFunctionalists can respond by widening the relevant system of relations. A belief’s role might include causal connections to substances, objects, social practices, and linguistic authorities beyond the organism. Alternatively, a theory may distinguish narrow content, associated with internal cognitive organization, from wide content, which includes environmental or social relations. The first approach expands functional individuation; the second limits what an internal functional theory is intended to explain. Neither response is cost-free. Wide individuation complicates the boundaries of psychological systems, while narrow content requires an account of what remains genuinely semantic after external reference is set aside.\n\nComputational approaches face the related symbol grounding problem: how do formally manipulated symbols acquire significance rather than merely standing in relations to other symbols? Some theories supplement functional organization with perception, action, learning, or teleosemantics, which connects representational content to biological functions and histories of selection. These proposals show why functionalism and a complete theory of content should not be treated as identical. A theory may successfully describe the causal processing of representations while still requiring an additional explanation of what makes those representations about particular features of the world."
    },
    {
      "heading": "Consciousness, qualia, and the explanatory gap",
      "text": "The most persistent objections to functionalism concern consciousness, particularly the felt character of experience. Seeing a vivid red surface, tasting something bitter, and feeling a sharp pain involve distinctive experiential qualities, often called qualia. Critics argue that a specification of inputs, internal transitions, and outputs might leave these qualities undetermined. A complete account of what a state does would then fail to capture what undergoing it is like. Functionalists must either explain phenomenal character through role or show that the apparent gap does not establish a difference in properties.\n\nThe inverted-spectrum argument asks whether two individuals could have systematically different color experiences while sharing the same functional organization. One person’s experience when viewing ripe tomatoes might resemble another person’s experience when viewing grass, although both make the same discriminations and use color words identically. The challenge requires more than a casual exchange of color labels. Actual color experience has a rich structure of similarity, saturation, brightness, memory, and association. A defensible inversion must preserve all functionally relevant relations, not merely a few verbal reports. Whether such a complete inversion is possible remains disputed.\n\nThe absent-qualia objection is more radical: a system might instantiate the relevant organization while having no experience at all. The idea is related to the philosophical zombie, though the arguments differ in scope. An absent-qualia case can target functional duplication alone, whereas a zombie is often described as a complete physical duplicate without consciousness. Ned Block’s “Troubles with Functionalism” (1978) developed influential objections suggesting that functional criteria could include systems whose possession of experience seems doubtful. Functionalists challenge either the coherence of these cases or the evidential force of the intuitions they elicit.\n\nThe knowledge argument, presented by Frank Jackson in 1982, raises a related issue. Mary, a scientist who knows all the relevant physical information about color vision while confined to a black-and-white environment, appears to learn something when she first sees color. If complete functional information were included in her prior knowledge, the case would also challenge the sufficiency of that information for understanding phenomenal character. Responses include the claims that she gains an ability, acquaintance, or a new way of representing an already known fact rather than knowledge of an additional nonphysical fact.\n\nThese disputes concern the explanatory gap between third-person accounts of organization and first-person experience. Some functionalists distinguish access consciousness—the availability of information for reasoning, report, and control—from phenomenal consciousness, the presence of experience itself. Critics argue that explaining access leaves the latter untouched. Defenders may argue that sufficiently detailed accounts of attention, discrimination, memory, and self-representation explain why experience appears resistant to objective description. The debate is not settled simply by showing that consciousness has functional consequences; the contested claim is that those consequences and their organization fully constitute it."
    },
    {
      "heading": "Chinese rooms, large systems, and implementation objections",
      "text": "The Chinese room argument, introduced by John Searle in “Minds, Brains, and Programs” (1980), challenges the sufficiency of formal symbol manipulation for understanding. A person who does not understand Chinese follows instructions for manipulating Chinese inscriptions and produces responses indistinguishable from those of a competent speaker. Searle argues that executing the program does not give the person an understanding of Chinese. The intended conclusion is that implementing a computer program, by itself, is insufficient for mentality; syntax is not thereby semantics.\n\nThe argument bears directly on computational versions of functionalism, but its relation to functionalism as a whole is less straightforward. A functional theory may include perception, action, learning, or causal relations that are not captured by a detached symbol-manipulation description. The “systems reply” also denies that the understanding, if present, must belong to the individual instruction-follower. It may belong to the organized system comprising the person, rules, records, and operations. Searle’s reply asks whether internalizing all these components would make the person understand. The dispute turns partly on which system is the appropriate bearer of mental properties.\n\nBlock’s large-system cases challenge functionalism from another direction. In one example, an enormous population coordinates its members to reproduce the causal organization attributed to a mind, communicating in ways that simulate its internal transitions. The example is often called the China-brain case. Its force is not that such a system violates the functional specification, but that the specification might be satisfied without any accompanying experience. A functionalist may accept the unexpected attribution of mentality, reject the imagined equivalence, or argue that intuitions about extraordinarily unfamiliar systems are unreliable.\n\nImplementation itself raises an important technical issue. If arbitrary correspondences are allowed between physical states and computational states, a simple physical object may be describable as implementing a surprisingly complex computation. An account that treats every convenient mapping as genuine implementation risks making computation—and consequently mentality—far too widespread. Restrictive accounts therefore appeal to the physical system’s actual causal structure, its capacity to support the relevant counterfactual transitions, and the organization of identifiable components. Merely matching a finite sequence of observed events is insufficient.\n\nThese objections expose a recurring tension between liberalism and exclusion. Functionalism should not exclude a genuinely minded system merely because its material or scale is unfamiliar. Yet an excessively coarse description may attribute mentality to collections, simulations, or accidental patterns that lack the relevant organization. The challenge is to specify principled conditions for realization without inserting unsupported requirements such as “must be made of neurons.” Whether computation, causal topology, biological organization, or interaction with an environment supplies those conditions remains a major division among theories."
    },
    {
      "heading": "Mental causation and physical explanation",
      "text": "Functionalism is often presented as explaining how mental states can matter causally: they are identified through their contributions to the production of thought and action. Yet this apparent advantage generates a problem when functional roles are distinguished from their physical realizers. If a particular neural event already has sufficient physical causes, what additional causal work remains for the higher-level property of being a belief or desire? The causal exclusion problem, associated especially with Jaegwon Kim, challenges theories that regard mental properties as distinct from their physical bases while also assigning them causal efficacy.\n\nThe problem can be illustrated by someone opening a window because they believe the room is stuffy and want fresh air. A psychological explanation cites the belief and desire. A physical explanation cites neural activity, muscle contraction, and bodily movement. Functionalists generally do not regard these as two independent chains that converge by coincidence. Nevertheless, if the mental property is distinct from the physical property that realizes it, critics ask whether the mental property is redundant, systematically overdetermining the outcome, or causally ineffective. The last possibility approaches epiphenomenalism, according to which mental phenomena do not influence physical events.\n\nRealizer functionalism offers one response: the causally effective mental state is the physical state occupying the relevant role. There is no additional event competing with it. Role functionalists may instead argue that higher-level properties capture patterns of dependence that physical descriptions obscure. The fact that a material is elastic can explain its response to a force without adding a second force to those described by mechanics. Similarly, a belief may explain behavior by identifying an organized causal capacity rather than an extra ingredient in the underlying process.\n\nInterventionist and counterfactual approaches to causation help formulate this second response. If appropriately changing a person’s belief changes their action across a range of underlying implementations, the belief variable may be causally informative. Such explanations select a stable level of description suited to the phenomenon being explained. Critics reply that explanatory usefulness does not automatically establish the independent causal efficacy of a higher-level property. The dispute therefore concerns both metaphysical dependence and the standards by which causal explanations are evaluated.\n\nFunctionalism also faces a question about causal specificity. A very broad role may classify together states with different detailed powers; an extremely narrow role may prevent different organisms or individuals from sharing any mental type. Psychological explanation needs categories general enough to support comparison but specific enough to support reliable dependencies. Research on learning, perception, and neural mechanisms can constrain these categories, although it does not by itself settle whether higher-level properties are identical with, constituted by, or otherwise dependent on their physical realizers."
    },
    {
      "heading": "Cognitive science, artificial intelligence, and embodied approaches",
      "text": "Functionalism helped establish the conceptual setting for cognitive science by legitimizing explanations in terms of internal organization between stimulus and response. The computational theory of mind develops one especially influential version: cognitive processes are computations over representations, and psychological explanation identifies the relevant representations and operations. Functionalism is broader than this theory. A system could be functionally described using continuous dynamics or distributed processes without treating every mental state as a discrete symbol in a classical program.\n\nDavid Marr’s Vision (1982) distinguished computational, algorithmic, and implementational levels of analysis. At the computational level, an explanation identifies the task and why it is appropriate; at the algorithmic level, it identifies representations and procedures; at the implementational level, it identifies the physical realization. This framework is not itself a complete functionalist metaphysics, but it illustrates why organization and material implementation can be investigated separately without being treated as unrelated. A theory of visual computation must still be assessed against what an actual visual system can implement.\n\nThe development of artificial neural networks further broadened the discussion. Distributed representations, graded activation, and learning through changes in connection strength differ from simple finite-state pictures or explicitly programmed symbolic rules. Nevertheless, these systems remain describable in terms of organized causal dependencies. The philosophical question is not whether they resemble the earliest machine models, but which features of their organization would be sufficient for cognition, understanding, or experience. Successful task performance establishes particular capacities; it does not by itself resolve every question about mental-state attribution.\n\nEmbodied cognition challenges approaches that treat the mind as an internally self-contained program. Bodily morphology, sensorimotor skills, and ongoing environmental interaction can shape what a cognitive system does and how it does it. A robot may exploit the passive dynamics of its limbs rather than explicitly calculate every movement. Such cases suggest that an adequate functional description sometimes needs to include body–environment coupling rather than a narrow sequence of internal computations. Some embodied theories reject functionalism, while others reinterpret functional organization at this wider scale.\n\nContemporary debates about large language models sharpen the distinction between behavioral evidence and constitutive theory. Linguistically appropriate responses can provide evidence of sophisticated information processing, but the significance of that evidence depends on which capacities are being assessed and which theory of mind is adopted. Different functionalisms require different combinations of internal organization, learning history, perceptual grounding, memory, and action. An assessment based solely on conversational output therefore cannot establish satisfaction of every functionalist account.\n\nScientific work continues to make these disagreements more precise by identifying mechanisms, testing models, and comparing alternative architectures. A functional characterization can generate hypotheses about which interventions should alter a capacity and which implementation changes should preserve it. The resulting evidence constrains proposed mental roles and their realizations, while leaving open the further philosophical question of whether those roles exhaust the nature of thought and experience."
    }
  ],
  "links": [
    "philosophy-of-mind",
    "physicalism",
    "artificial-intelligence",
    "belief",
    "counterfactual",
    "analytic-philosophy",
    "behaviorism",
    "alan-turing",
    "david-lewis",
    "turing-machine",
    "intentionality",
    "consciousness",
    "qualia",
    "chinese-room",
    "john-searle",
    "causation",
    "cognitive-science",
    "artificial-neural-network",
    "embodied-cognition",
    "large-language-model"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}