Distributed Cognition

Two distinct research programmes travel under this heading, and merging them is the commonest error in the area. The first is Edwin Hutchins’ ethnography of real work: Cognition in the Wild (1995) treats a naval navigation team as the computational unit, tracking how representational state is propagated and transformed across charts, instruments, telephone circuits and people. The second is Andy Clark and David Chalmers’ extended mind thesis, published in Analysis in 1998, which argues from a parity principle that an external resource playing the functional role an internal one would play is literally part of the cognitive system. Hutchins is an anthropologist choosing a unit of analysis. Clark and Chalmers are philosophers making a claim about constitution. Different founders, different methods, different journals, different commitments — and only one of the two needs any metaphysics at all.

This brief keeps them apart, and then makes an argument about the second that separates this topic from everything else in Category IV. In Consciousness Research, in Integrated Information Theory, in Machine Consciousness, the blocker is that we cannot yet measure the thing. Here the blocker is a different kind of obstacle. The extended hypothesis and its principal rival — Robert Rupert’s claim that cognition is deeply embedded in the environment without being constituted by it — may make identical predictions about everything that can be measured. If they do, the dispute is not empirical, and no experiment retires it. Rupert’s own observation that internal and external memory differ in their dispositional profiles is the only concrete divergent prediction this brief could locate, and it appears not to have been tested. That untested experiment is the spine of what follows: sections 4 and 6 are built around whether this is a scientific question at all, and around the specific study that would settle whether it is.

Established As a method, distributed cognition works, and that part is not in dispute. Hutchins’ Cognition in the Wild (MIT Press, 1995) analysed the navigation team of the USS Palau as a system in which mental representations, which classical cognitive science placed inside the individual brain, are instead distributed across a sociocultural system of people and artefacts. Established The companion study, “How a Cockpit Remembers Its Speeds”, Cognitive Science 19(3):265–288, 1995, made the same move for aviation: the speed bugs on an airspeed indicator are not a memory aid used by a pilot who remembers, they are where part of the remembering happens. Established The methodological core is a choice of unit — on the Rogers and Ellis (1994) formulation, the right object of analysis is a collection of individuals and artefacts and their relations to each other in a particular work practice. Frontier That choice is defensible entirely on explanatory grounds, without any claim that the cockpit has beliefs, and it is the reason the programme has thrived. Established Hollan, Hutchins and Kirsh, in ACM Transactions on Computer-Human Interaction 7(2):174–196 (2000), carried it into design practice with an emphasis on coordination between internal and external structure; Frontier distributed cognition consequently has far more institutional traction in human-computer interaction, robotics and design than it has ever had in philosophy of mind. Established A provenance caution: this brief reached the Hutchins literature through a secondary source rather than the primary texts, and the bibliographic details, though standard, were not independently confirmed.

Established The extended mind thesis is a much stronger claim, and it is a claim about constitution rather than about explanation. A. Clark and D. Chalmers, “The Extended Mind”, Analysis 58(1):7–19, 1 January 1998, DOI 10.1093/analys/58.1.7 — a record confirmed by direct registry lookup for this brief. Established Its engine is the parity principle: if a process in the world functions as a process in the head would, and we would count the in-head process as cognitive, then the external process is cognitive too. Frontier A bibliographic warning attaches to that sentence: the encyclopaedic source available to this brief does not name the parity principle, and the formulation, though standard in the literature, should be quoted from the 1998 paper, which this brief did not obtain. Established The illustration is Otto and Inga. Both travel to a museum; Inga recalls the address from biological memory, and Otto, who has Alzheimer’s disease, consults a notebook he always carries. The claim is that Otto’s belief about the museum’s location is partly constituted by the notebook.

Established The thesis is guarded by conditions on what counts as an Otto notebook, and this brief can only report three of them. The recovered conditions are that the resource is constantly and immediately accessible, that it is automatically endorsed by the user, and that it functions with the same purpose as the internal process. Frontier The canonical fourth condition — that the information was consciously endorsed at some point in the past — was not recovered from any source available to this brief, and the conditions are usually given as four. Established This brief therefore does not present a complete condition list. That matters more than a bibliographic footnote would suggest, because the conditions are the only thing standing between the thesis and the objection that follows.

Established The principal objection is Fred Adams and Kenneth Aizawa’s, and it dates from 2001, not 2010. “The bounds of cognition”, Philosophical Psychology 14(1):43–64, March 2001, DOI 10.1080/09515080120033571 — a verified registry record. Established The widely cited 2010 items are the book The Bounds of Cognition and a chapter in Richard Menary’s anthology; the most-read encyclopaedic summary dates the objection to 2010 and thereby loses nine years of priority. Established The objection has two separable parts. The first is the mark of the cognitive: cognition has a distinguishing feature, for Adams and Aizawa non-derived or intrinsic content, and external representations such as notebook entries carry only derived content, so they are not cognitive. Established The second is the coupling-constitution fallacy: from the fact that something is causally coupled to a cognitive system it does not follow that it is part of one. Their reductio, as quoted in the source this brief could read, is the question of why the pencil thought that two plus two equals four — because it was coupled to the mathematician. Established They add the cognitive bloat worry: the thesis appears to have no principled stopping point short of absorbing other people, institutions and the internet.

Established Rupert’s objection is a different one and is routinely conflated with Adams and Aizawa’s. R. D. Rupert, “Challenges to the Hypothesis of Extended Cognition”, The Journal of Philosophy 101(8):389–428, 2004, DOI 10.5840/jphil2004101826. Established He does not argue that extended cognition is incoherent. He argues that it is explanatorily inferior: a science of the embedded mind, in which cognition depends deeply on environmental scaffolding without being constituted by it, explains everything the extended thesis explains with fewer commitments. Frontier His second and more consequential point is empirical in form. Otto’s notebook-belief and Inga’s memory-belief differ in their dispositional profiles — interference effects, generation effects, recall dynamics — in ways that matter to cognitive science, and a theory that treats them as the same kind of state throws that structure away. Established This is the default position of most working cognitive scientists and is the mainstream, not a fringe complaint. Frontier Rupert also wrote the field’s reference surveys — “Situated Cognition” for Oxford Bibliographies in Philosophy (2018) and “Embodied cognition” for the Routledge Encyclopedia of Philosophy (2019) — and a leading critic writing the discipline’s reference articles is a useful signal about where the centre of gravity sits.

Frontier Clark’s reply concedes more than it appears to. In Supersizing the Mind (2008, p. 96) the claim is reframed: not that the processes in Otto and Inga are identical, but that information is poised to guide reasoning and the behaviour that counts. Frontier Critics read that as a retreat from parity to functional poise, and therefore as conceding Rupert’s point — if the processes need not be alike, the argument for treating the notebook as a constituent rather than a cause has lost its engine. Speculative A defender can reply that constitution never required process identity, only the right functional role. Handwave This brief declares the tie: the exchange is a dispute about what parity was ever supposed to establish, and neither side has produced an observation that bears on it.

Established The wider 4E framework — embodied, embedded, enacted, extended — is a family, not a theory. Its shared commitment is that cognition is shaped and structured by dynamic interactions between brain, body, and physical and social environments. Established The founding enactivist text is F. Varela, E. Thompson and E. Rosch, The Embodied Mind (1991), whose definition holds that cognition depends on the kinds of experience that come from having a body with sensorimotor capacities, and that those capacities are themselves embedded in a biological, psychological and cultural context. Established Margaret Wilson’s 2002 taxonomy is the standard map: cognition is situated; cognition is time-pressured; we off-load cognitive work onto the environment; the environment is part of the cognitive system; cognition is for action; off-line cognition is bodily-based. Established Wilson herself flags the fourth claim as deeply problematic, on the ground that not everything with an impact on the elements of a system is part of the system. Frontier That is the coupling-constitution objection raised from inside the programme, in its most-cited taxonomy paper, against the one claim of the six that the extended mind thesis needs. It is an under-used fact and this brief treats it as decisive about the internal state of the field. Established The pre-history is J. J. Gibson’s ecological psychology and affordances; Established the linguistic strand is Lakoff and Johnson’s conceptual metaphor theory; Established the perceptual strand is sensorimotor contingency theory, developed by J. K. O’Regan and Alva Noë by the early 2000s and set out in Noë’s Action in Perception (MIT Press, 2004). Frontier The full citation for the O’Regan and Noë journal paper was not recovered for this brief and is not stated here.

Frontier The embodiment evidence base splits into two halves with very different standing, and this brief cannot report on one of them. On one side sit neural results: functional imaging reported in 2004 in which passively reading action words such as lick, pick and kick produced somatotopic activity in or adjacent to regions associated with the corresponding movement; a 2005 transcranial-magnetic-stimulation result coupling motor system activity to auditory action sentences; electrophysiological findings that sensory and motor processes are strongly coupled rather than sequentially separated. Frontier The source available to this brief gave no author attribution for either the imaging or the stimulation study, so they are described here and cited nowhere. Frontier On the other side sit behavioural priming results: objects judged heavier when described as important, distances judged further by people in chronic pain, autobiographical recall speeded by compatible body postures, and approach-avoidance effects for positive and negative words. Speculative This second class is where the replication problems live, and this brief states plainly that it could not obtain the replication literature. The encyclopaedic source that supplied the list above was explicitly assessed and found to contain no mention of failed replications or of the embodiment priming crisis at all. Handwave No replication rate appears anywhere in this brief, because a number stated without its evidence would be worse than the gap. Section 14 returns to this as the topic’s largest single defect.

Frontier The first frontier question is whether there is a frontier. The critique has institutionalised: Rupert’s two reference surveys (2018, 2019) mark the point at which the disagreement stopped being a controversy and became a map. Frontier That is the news. A debate that stabilises without resolving, in which neither side expects data to arrive, is behaving like a dispute about how to describe a domain rather than what is in it. Speculative The optimistic reading: this is what a mature programme looks like before its decisive experiment. Handwave The pessimistic reading: no such experiment is coming, because none exists.

Frontier The live methodological crisis is in embodiment, not in extension, and this brief’s sources were blind to it. The primary replication literature — the multi-laboratory projects, the registered replication report on facial feedback, the record on power posing and elderly priming — was not obtained for this brief, and the encyclopaedic article that supplied its embodiment evidence base omits the subject entirely. Speculative The likeliest state of the field, on this brief’s reading, is that the neural and stimulation evidence survives while a substantial fraction of the behavioural priming evidence does not — leaving embodiment intact as a claim about sensorimotor grounding while it loses its best-known results. Handwave That is a judgement, not a finding, and cannot become one without the literature this brief lacks.

Frontier Predictive processing has partly absorbed the embodiment programme, and nobody agrees what that means. Seth and Bayne’s four-family survey (Nature Reviews Neuroscience 23:439–452, 2022) treats re-entry and predictive processing as one family, and interoceptive versions make bodily regulation constitutive of selfhood. Frontier Vindication of embodied cognition, or recapture by computationalism? The predictive brain is a body-centred brain; it is also an inference machine running generative models, which is the representational apparatus radical enactivism exists to eliminate. Speculative The question is open and the answer is consequential, because predictive processing is where the funding is.

Frontier The programme’s most consequential current export is an argument against machine consciousness. Aru, Larkum and Shine argue that large language model inputs lack the embodied, embedded information content characteristic of human sensory contact with the world. Frontier This is the clearest link between this brief and Machine Consciousness, and a programme whose own metaphysics is contested is supplying the sharpest premise in a neighbouring debate. Established The claim is specific: it concerns the information content of training data, not a general theorem that minds need bodies. Speculative If right, embodiment is a bottleneck for artificial consciousness that scaling does not relieve; Handwave if the sensorimotor strand is right instead, a robot with real sensorimotor mastery clears the bar and the objection expires.

Speculative The exotic end of the topic is the version its founders declined to defend. Clark and Chalmers extended belief; they did not extend experience, and Clark has resisted doing so. Speculative Extended consciousness would require vehicle externalism for phenomenal states, not merely for content; Frontier it would need a story about how extra-neural structure contributes to the unity of an experience; Frontier and it faces a timescale obstacle, since the neural processes associated with conscious access run at tens to hundreds of milliseconds while consulting a notebook takes seconds. Speculative That is where extended consciousness actually fails, not merely where it seems implausible.

Frontier Group minds embrace the bloat rather than resisting it. Extending the unit-of-analysis move to organisations, markets and scientific communities requires only that the coupling-constitution objection fail at social scale, and the socially extended literature treats absorbing other people as a feature. Speculative Autopoietic enactivism reaches the same territory from the other direction: if any operationally closed, self-producing system has cognitive capacities, cognition is co-extensive with life and bacteria and plants cognise. Frontier That life-mind continuity thesis intersects the New York Declaration of April 2024 and its 500-plus signatories — on the continuity view, extending the circle to insects is unremarkable; on almost every other view it is startling. Handwave Radical enactivism sits at the far end: Hutto and Myin hold that basic minds have no content at all. It is the only live position in cognitive science that treats representation as the thing to be eliminated rather than assumed, and its exposed flank is the scale-up objection.

Frontier A caution about this section. The research behind this brief had no working preprint search and lost two bibliographic services. Speculative The frontier layer is under-sampled by construction; assume recent work exists that this brief did not see.

Frontier The binding constraint is not a measurement problem. It is the question of whether there is anything to measure. Constitution is not directly observable. Every experiment showing that external resources matter to cognitive performance is equally well explained by the embedded reading, on which the resource is a cause of cognitive states rather than a part of one. Speculative Until someone derives a prediction on which the extended and embedded hypotheses differ, the dispute has no empirical content, and the honest description of the field is that its central claim may be idle. Handwave Everywhere else in this category the obstacle is instrumentation; here it is that the two rivals may agree about every observation.

Frontier The second bottleneck is the mark of the cognitive, and it is not an experimental question. Adams and Aizawa’s demand that cognition involve non-derived content cannot be adjudicated by any single measurement; it requires a naturalistic account of how content arises at all, which is the same problem Hutto and Myin call the hard problem of content and declare unsolvable. Speculative A programme that settled content naturalisation would decide the extended mind debate as a corollary. Handwave No such programme has a completion criterion, and none of the parties has proposed one.

Frontier The third bottleneck is timescale, and it binds only the exotic version. If extended consciousness is to be defended, someone must show that processes operating at the timescale of reaching for a notebook can contribute to states with the temporal signature of conscious access. Speculative This is a specific and falsifiable-looking obstacle rather than a vague implausibility, which makes it the most interesting negative result available anywhere in this topic.

Established The fourth bottleneck is institutional and has no scientific component whatever. Recovering the embodiment evidence base requires systematic pre-registered replication of the behavioural priming results, cleanly separated from the neural and stimulation results, which have different evidentiary standing and should not stand or fall together. Established Nothing about this is hard. Frontier It is unglamorous, it is under-funded, and it is nobody’s career, which is a market failure rather than a scientific barrier — and it is the item on this list with the highest ratio of value to difficulty.

Frontier The fifth is industrial and is already partly cleared. Demonstrating that the distributed-cognition unit of analysis yields design predictions the individualist frame does not is a matter of building systems and testing them, which human-computer interaction has been doing since the Hollan, Hutchins and Kirsh synthesis. Speculative It is worth noting what this bottleneck does and does not settle: design wins support distributed cognition as a method and are silent on extension as a metaphysics.

Frontier This brief waits on results produced elsewhere in the register. The timescale argument against extended consciousness borrows its numbers from the empirical programme surveyed in Consciousness Research: the temporal signature of conscious access, and the degree to which that signature is theory-neutral rather than an artefact of one framework’s preferred paradigm. Frontier If those figures move, the argument moves with them.

Frontier It depends second on the memory sciences, which it has barely drawn on. The only concrete divergent prediction in this area is a claim about dispositional profiles — interference, generation effects, retrieval dynamics — and testing it requires the standard laboratory paradigms of human memory research applied to externally stored information. Established Those paradigms exist and are mature. Speculative The dependency is not on a new method but on somebody deciding the question is worth their apparatus.

Frontier Third, it depends on a naturalistic theory of content for the reasons in section 4, and on the philosophical programmes that would supply one. Handwave This is the dependency with no delivery date, and the one that would settle the argument outright if it ever arrived.

Established Fourth, it depends on functioning replication infrastructure — multi-laboratory consortia, registered reports, and journals that will publish a well-powered null. Frontier The relevant fact for this brief is that the infrastructure exists in psychology and has not been pointed at the embodiment corpus in any way this brief could confirm.

Frontier Fifth, it depends on access. The research behind this brief could not reach the full text of the founding 1998 paper, could not reach the replication literature, and lost two bibliographic APIs entirely. Established That is a dependency on scholarly access arrangements, and it is a real constraint on what any brief in this area can responsibly assert.

Frontier The experiment that would make this dispute empirical is available now and appears not to have been run. Rupert’s claim is that notebook-belief and memory-belief differ in dispositional profile. That is testable. The design: a longitudinal study of heavy external-memory users — dense digital note-takers, and amnesic patients using prosthetic memory systems — probing whether externally stored information carries the signatures of internally stored information. Established The signatures are standard: interference from similar items, the generation effect, retrieval-induced forgetting, and the recall dynamics that distinguish recollection from familiarity. Frontier The predictions diverge in a stateable way. On a strong extended reading, an item that satisfies the trust-and-glue conditions should behave like a remembered item on at least some of these measures. On the embedded reading it should not, because it is not a memory at all but a fluently accessed external record.

Speculative Both possible outcomes are informative, which is rare in this area. If external items show the dispositional signatures, the extended theorist has the first empirical purchase the thesis has ever had. If they do not, the extended theorist must either concede Rupert’s point or retreat to a functional-poise claim that has no further predictions — and the retreat is itself a result. Handwave The failure mode to guard against is a design in which the extended reading can absorb either outcome, which is the pattern that has kept this argument alive for twenty-five years.

Established The second experiment is the replication programme. Pre-registered, adequately powered, multi-site replication of the embodiment priming corpus, with the neural and stimulation results held separately and replicated on their own terms. Frontier This brief cannot say what such a programme would find, and says so rather than guessing.

Speculative The third is the timescale test. Instrument a task in which an external resource is consulted within the window associated with conscious access rather than seconds outside it — a heads-up display, a haptic channel, a closed-loop interface — and ask whether anything about the resulting experience differs from the slow-consultation case in a way the extended theorist predicted in advance. Handwave No such prediction has been registered, which is the honest state of play.

Frontier The instrumentation this topic needs is unglamorous and mostly exists. Testing dispositional profiles in the wild requires logging real external-memory use at fine grain — what was written, when it was consulted, what was retrieved without consultation — over months rather than in a single laboratory session. Established Note-taking systems already produce this telemetry as a by-product. Frontier The engineering problem is not capture but consent, retention and the construction of a research instrument out of a commercial product without inheriting its design decisions — a note-taking application optimised for retrieval convenience has already made choices that bear on the very dispositions the study is trying to measure. Speculative A purpose-built instrument, deployed to a cohort for a year, is a modest engineering project by the standards of this register and would produce the first longitudinal record of external-memory use anyone could analyse.

Established The sensorimotor strand has the most testable rigs in 4E cognition. Sensory-substitution devices — tactile or auditory encodings of visual scenes — make the enactivist claim that perceptual quality is a matter of mastery over sensorimotor contingencies directly assessable, because mastery can be trained and measured. Frontier They are also the natural bridge to Brain-Computer Interfaces, where the same question arises with a different transducer.

Frontier Prosthetic-memory systems are the strongest available Otto case and are engineering artefacts, not thought experiments. A well-designed memory prosthesis for amnesia satisfies the recovered trust-and-glue conditions better than any notebook: constantly accessible, automatically consulted, functionally continuous with what it replaces. Speculative Building one specifically as a test instrument — with the retrieval paradigms of section 6 designed in from the start rather than bolted on — is the single most useful engineering contribution anyone could make to this dispute. Handwave No group appears to be building one for that purpose.

Established Human-computer interaction is where this programme actually lives. The Hollan, Hutchins and Kirsh synthesis made distributed cognition a design methodology, and computer-supported cooperative work has used the unit-of-analysis move for two decades without needing to settle anything metaphysical. Frontier That divergence — institutional success in design, stalemate in philosophy — is the single most informative fact about the topic’s health.

Frontier Behaviour-based robotics is the engineering ancestor. The move away from central world-models toward tightly coupled sensorimotor loops was a working demonstration that a great deal of apparently cognitive behaviour needs no internal representation, and it shaped the philosophical programme rather than following it. Speculative Its limit is the scale-up objection, which robotics hit before philosophy formulated it.

Frontier Collective intelligence and organisational design inherit the socially extended claim. Collective Intelligence asks what makes a group perform above or below its members; distributed cognition supplies the vocabulary for answering it structurally rather than as a sum of individual competences. Speculative Transactive memory — the observation that teams divide not only what is remembered but the knowledge of who remembers what — is the most direct point of contact, and it is where cognitive offloading stops being a metaphor and becomes a measurable property of a working group. Frontier This brief flags that its research obtained no primary source for either the transactive-memory literature or the search-engine offloading studies, cites neither, and states none of their results; the gap is named here so that it is addressable rather than invisible.

Frontier The AI-facing neighbours are Human-AI Integration, Cognitive Architectures and Human Cognitive Augmentation. Speculative Each becomes a different question depending on whether an external resource participating in reasoning is a part of the thinker or a tool the thinker uses — and none of the three can wait for an answer, because the systems are already deployed.

Established The topic’s institutional pathology is that its two halves are refereed by different disciplines with different standards. Hutchins’ programme is judged by ethnographic and design criteria; the extended mind is judged by philosophical argument; the embodiment evidence base is judged by experimental psychology, which is the only one of the three with a replication culture. Frontier Claims migrate freely across the boundary and their evidentiary standing does not migrate with them. A priming result of uncertain standing is cited in a philosophy paper as though it were a fixed point, and a philosophical thesis is cited in a design paper as though it licensed a method it does not.

Established The replication of the embodiment corpus is an institutional problem with an obvious owner and no owner. Multi-laboratory consortia and registered reports are established instruments in psychology; the embodiment corpus is a well-defined target; the work is cheap by the standards of this register. Frontier What is missing is a funder who will pay for a systematic audit of a literature rather than for new results, and a hiring committee that will count one.

Frontier Access is an institutional constraint on this brief itself, and it should be recorded rather than hidden. The research behind this document could not retrieve the full text of the founding 1998 paper, lost two major bibliographic services outright, and had no working preprint keyword search. Established As a direct consequence, three of the four canonical conditions on Otto’s notebook are stated here and the fourth is not.

Frontier Two tooling failures during the same work are worth reporting as institutional facts about automated scholarship. A publisher URL served a different article than the identifier requested, and a bibliographic fuzzy matcher returned wholly unrelated papers with no error and no low-confidence signal. Established Either would have produced a confident false citation if unchecked. Speculative As briefs of this kind are increasingly assembled by machine, silent retrieval failure becomes a systematic risk to the record rather than an occasional nuisance, and the mitigation is procedural: check the identifier on the page you land on against the identifier you asked for, and compare a returned title to the query before recording anything.

Speculative If the extended thesis is true, it changes what counts as an injury. Destroying Otto’s notebook is, on the strong reading, closer to harming a person than to damaging their property, and the intuition that this is obviously wrong is doing real argumentative work for the critics. Frontier The bridge from constitution to moral status is not automatic and nobody has built it carefully, Speculative but the extended mind is one of the few places in this register where a metaphysical thesis has consequences that could in principle be litigated.

Frontier The practical version of that argument is already in circulation under other names. Cognitive Liberty and the neurorights literature argue for protections over mental states; if cognitive states extend into devices, the protected boundary moves outward to include the device, the notes and the query history. Speculative That is a stronger conclusion than most privacy frameworks currently reach, and it follows from a premise many of their proponents have not examined.

Frontier The AI case sharpens it. If a person reasoning with a model has an extended cognitive system, then the model’s provider holds part of the user’s cognition, with implications for consent, continuity of service and the ethics of unannounced change. Speculative If the person merely has a good tool, the ordinary framework of product responsibility applies and nothing exotic follows. Handwave Nothing currently distinguishes the two readings observationally, which means an ethical framework may have to be chosen before the metaphysics is settled — and this brief takes that to be the honest situation rather than a reason to wait.

Established There is a duty-of-care question inside the field as well. Embodiment priming results are taught, cited in applied settings and used to justify interventions; if a material fraction of them do not replicate, the field has an obligation to establish that rather than to leave it as a rumour. Frontier Not knowing is a choice at this point, not an accident.

Speculative The socially extended thesis, if taken seriously, relocates the unit of epistemic evaluation. Scientific communities, markets and bureaucracies would not merely support cognition but perform it, and the right question about an institution becomes what it can think rather than what its members believe. Frontier Hutchins’ unit-of-analysis move already licenses that question methodologically, without the metaphysics, which is why a research programme about ships and cockpits has more to say about institutional design than most theories of organisation do. Speculative The strong version goes further and treats the institution as the bearer of the cognitive state, which is where the coupling-constitution objection returns at social scale and where the socially extended literature mostly declines to resist it.

Frontier The civilizational stake is in the trajectory of offloading rather than in the philosophy. Populations are shifting an increasing share of memory, calculation and now inference into external systems, and the empirical questions that matter — what is retained, what is lost, what is never encoded, how skill degrades and how it recovers — are the same whether the external resource is a constituent or a cause. Speculative This is the point at which the metaphysical dispute becomes strategically irrelevant and the measurement problem becomes urgent, and it is an argument for funding the dispositional-profile work on grounds that have nothing to do with settling the philosophy.

Speculative The asymmetry worth naming is ownership. A notebook is owned by the person who carries it. A model is not. Handwave A civilization that has distributed a large fraction of its cognition into systems it does not control has made a structural commitment whose reversibility nobody has assessed, and the extended-mind literature, for all its metaphysical ambition, has almost nothing to say about it.

These horizons track the empirical question — whether the extended and embedded hypotheses can be told apart — rather than the philosophical literature, which has no schedule.

• 10 yr: Frontier The replication audit of the embodiment priming corpus is done or conspicuously not done; either way the field knows which of its public results survive. Speculative The dispositional-profile experiment is run at least once, most plausibly as a side-product of memory research on digital note-taking rather than as a test of the extended mind. Frontier Distributed cognition continues to consolidate in design and human-computer interaction without settling anything metaphysical. Speculative The transactive-memory and offloading literatures are re-audited under modern standards, because they are cited constantly in applied settings and their standing is not currently clear.
• 25 yr: Speculative Either a divergent prediction has been established and this becomes an ordinary empirical programme, or the field has converged on the embedded framing as the default and the extended thesis survives as a philosophical position with a small literature. Frontier This brief judges the second outcome more likely. Speculative The AI case forces a legal answer about ownership of externalised cognition well before the metaphysical one arrives, and the answer is given by courts and regulators rather than by cognitive science. Frontier Either way, the applied literature on offloading and team memory has by then become the part of this topic that matters practically.
• 50 yr: Speculative A mature science of human-machine cognitive systems exists and is indifferent to the constitution question, having found its own units of analysis. Handwave If content naturalisation is solved in the interim, the mark-of-the-cognitive dispute resolves as a corollary and the 1998 argument gets a verdict rather than a stalemate. Speculative The legal system, meanwhile, will have decided several times over who owns an externalised cognitive state, on grounds that owe nothing to the philosophical literature.
• 100 / 250+ yr: Handwave Extended consciousness — the version Clark and Chalmers declined to defend — is either demonstrated by interfaces fast enough to close the timescale gap, or it is retired as the clearest example in this register of a hypothesis that was never empirical. Handwave The second is the way to bet.

• Depends on This brief depends on the empirical programme surveyed in Consciousness Research for the one figure its exotic version turns on: the temporal signature of conscious access. The argument that experience cannot have extra-neural vehicles is a timescale argument, and it inherits both that figure and every uncertainty attached to it — if the window is wider than currently reported, or if it is an artefact of the report-based paradigms that dominate the field, the objection weakens accordingly. The dependency is therefore on a result that is itself contested, which is worth stating rather than assuming away. It depends second on the memory sciences, which own the paradigms — interference, generation effects, retrieval-induced forgetting — that supply the only concrete divergent prediction anybody has proposed, and which have never been applied to the question. It depends third on a naturalistic theory of mental content, which would settle the mark-of-the-cognitive objection as a corollary and which has no delivery date and no completion criterion. And it depends fourth on replication infrastructure in experimental psychology: multi-site consortia, registered reports, and journals that will publish a well-powered null. That infrastructure exists. It has not been pointed at the embodiment corpus.
• Enables What this brief enables is mostly vocabulary rather than capability, and that is an honest description rather than a modest one. Distributed cognition supplies the unit of analysis that collective-intelligence work needs to explain group performance structurally rather than as a sum of individual competences, and that human-computer interaction has used as a design methodology for two decades without waiting for the metaphysics. The embodiment argument supplies the sharpest current premise in the case against machine consciousness — that model inputs lack the embodied, embedded information content of sensory contact with a world — and that premise is now doing more work in AI assessment than it ever did in philosophy of mind. The extended thesis, if it were ever established, would enable a considerably stronger neurorights claim than current privacy frameworks reach, by moving the protected boundary of the mind outward to include the device, the notes and the query history. And the negative result would enable something too: a field that established its central dispute to be non-empirical would have retired a question rather than lost one.
• Adjacent Collective Intelligence, Human-AI Integration, Cognitive Architectures, Brain-Computer Interfaces, Cognitive Liberty, Machine Consciousness, Human Cognitive Augmentation. The adjacency that matters most is the first: collective intelligence asks empirically what the socially extended mind asks metaphysically, and it has data.

Established “The Adams and Aizawa objection dates from 2010.” Wrong, and the error is in the most-read reference source on the subject. The objection is “The bounds of cognition”, Philosophical Psychology 14(1):43–64, 2001, DOI 10.1080/09515080120033571 — a registry record verified for this brief. Established The 2010 items are the book and the Menary-volume chapter. Frontier Nine years of priority are lost by the misdating, and with them the fact that the principal objection is nearly contemporaneous with the thesis it attacks.

Established “Rupert argues that extended cognition is incoherent.” Wrong. He argues it is explanatorily inferior. Established The 2004 paper defends the embedded alternative on grounds of parsimony and of better fit to the dispositional differences between internal and external memory. Frontier Presenting him as an in-principle objector misstates the most influential critique in the literature and hides the one testable claim in it.

Established “Distributed cognition and the extended mind are the same thesis.” Wrong, and this is the error that does the most damage. Established Hutchins’ distributed cognition is an ethnographic and methodological programme about units of analysis; Clark and Chalmers’ extended mind is a metaphysical thesis about constitution. Frontier Distributed cognition needs no metaphysics at all, which is precisely why it succeeded in design while the extended mind stalled in philosophy. Speculative Citing Hutchins’ fieldwork as evidence for the extended mind is the standard form of the mistake: an ethnography of how a navigation team works does not establish that its instruments constitute anybody’s beliefs.

Speculative “Embodied cognition is well replicated.” Not established, and there is reason for concern. Frontier The neural and stimulation results and the behavioural priming results have different evidentiary standing and are routinely cited as one body of evidence. Established The encyclopaedic source this brief relied on was checked and contains no mention of failed replications or of the priming crisis. Handwave This brief therefore prints no replication rate, and names the absence as its largest single defect. A brief that guessed a number here would read more comfortably and be less useful.

Established “Clark and Chalmers claimed that conscious experience extends beyond the brain.” Wrong. They extended dispositional belief. Established Clark has resisted extending experience, and the Supersizing the Mind formulation frames the claim as information poised to guide reasoning rather than as process identity. Speculative Extended consciousness is a distinct and far less defended position, Frontier and it is the one this brief regards as most likely to be decisively refutable, on timescale grounds.

Established “Embodied cognition researchers accept that the environment is part of the cognitive system.” Not uniformly, and the doubt comes from inside. Established Wilson’s own 2002 taxonomy flags that claim as deeply problematic, on the ground that not everything affecting a system’s elements belongs to the system. Frontier It is the least accepted of her six claims and the one the extended mind requires.

Established “The four Es are a unified alternative to cognitivism.” Wrong. Established Extended cognition is functionalist and representation-friendly; radical enactivism eliminates representational content; Clark defends extension while attacking sensorimotor enactivism. Frontier They share an enemy, not a theory, and treating the letters as a programme obscures that the positive claims may be mutually inconsistent.

Established “Otto’s notebook counts as memory because it is convenient.” Understates the conditions. Established The thesis requires constant accessibility, automatic endorsement and functional equivalence of role — conditions designed precisely to block cognitive bloat. Frontier This brief recovered three of the four usually given; the missing condition concerns prior conscious endorsement, and its absence weakens exactly the defence that is supposed to keep the internet out of your mind.

Established “Enactivism means cognition requires a body, so AI cannot cognise.” Conflates strands. Established Autopoietic enactivism makes cognition co-extensive with life; sensorimotor enactivism makes perceptual quality a matter of sensorimotor mastery, which a robot could in principle have; radical enactivism is a thesis about content, not about bodies. Frontier The embodiment argument against machine consciousness is a specific claim about the information content of training data, not a general enactivist theorem.

Handwave “The extended mind is a substantive empirical claim that will eventually be settled by neuroscience.” This brief’s central negative finding is that this may be false. Speculative No result in cognitive neuroscience distinguishes a resource that constitutes a cognitive state from one that merely causes it, because constitution is not a thing detectors detect. Frontier The single candidate for a divergent prediction is Rupert’s dispositional-profile claim, and it appears not to have been tested. Handwave If it is tested and the profiles are indistinguishable, the correct conclusion is not that the extended thesis is false but that it was never a hypothesis — and a field willing to say that about itself would be in better shape than one that is not.

Frontier A note on the sourcing of this brief, offered as a correction to itself. The work behind it had no functioning preprint search and lost two bibliographic services entirely; a publisher URL served a different article than the identifier requested, and a fuzzy bibliographic matcher returned wholly unrelated papers with no error signal. Established Several sources below are verified bibliographic records whose text was not read, and are marked as such rather than dressed up. Speculative The claim to make about a brief assembled under those conditions is not that it is complete but that it is honest about where it is not.