{"id":"87ae99fc-4ba9-4ecf-b2c7-845944bf9325","arxiv_id":"cs/0603125","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Computation is defined as using contingent environmental processes to externalize thought, and the Church-Turing thesis is recast as a demand that rigorous thought be expressible as software.","lead":"This paper argues that a physical process is a computation only when someone uses it to externalize thought. It reinterprets the Church-Turing thesis as a claim about rigorous thought, not about which functions are computable.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The fixed/contingent split that anchors the definition of computation is never formally defined; since the split is user-relative, the central claim becomes vacuous and unfalsifiable.","rationale":"The reader's weakest assumption was the anti-Platonist premise that ideas exist only when someone is thinking them. That is a genuine metaphysical commitment, but the paper explicitly labels it a stance and says 'these issues can be worked out.' Moreover, the central definition as 'externalized thought' could in principle survive a Platonist reading if 'use' remains essential. The more load-bearing gap is the asserted fixed/contingent separation, which is the mechanism meant to give computation a non-vacuous structural content. Without a principled criterion for drawing the boundary, the definition is underdetermined: a process can be partitioned in multiple ways, each yielding a different 'computation,' and the paper provides no way to adjudicate. The self-modifying-code check exposes this ambiguity directly, since it involves a case in which the program—which the paper places in the contingent environment—can change itself, making the fixed/contingent line shift during the computation. This does not change the reader's CONDITIONAL verdict, but it shifts the emphasis from the metaphysical premise to the analytical tool. The paper's historical narrative and the reformulation of the Church-Turing thesis remain thought-provoking, and the lack of formalization is appropriately reflected in a conditional acceptance.","tokens_in":10573,"tokens_out":7570,"duration_ms":72010,"concrete_test":"Test the paper's definition on a standard von Neumann self-modifying program. Following Section 3, take the CPU as the fixed process and the memory as the contingent environment. In self-modifying code, the resident program alters the instructions being executed. Determine, from the paper's text, whether that program is fixed or contingent: if it is contingent (because it lives in memory), then the 'computation' changes mid-execution; if it is fixed, then the split is no longer the CPU/memory boundary illustrated in the paper. The text provides no rule for this case. If both readings are admissible, the fixed/contingent distinction is underdetermined and the central definition fails to be well-defined.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—'A process in nature may be considered a computation only when we use it as a way to work with externalized thought'—relies on the asserted principle that 'we can always separate a computational process into its fixed part and its contingent or alterable part' (Section 3). This split is load-bearing because it converts the definition of computation from a purely mental act into something with an identifiable structural core. Yet no criterion for making the split is given. The examples are persuasive but circular: the CPU is called fixed and the memory contingent, but microcode is alterable; the balance scale is fixed and the objects loaded contingent, but the scale's own construction is contingent on its manufacturing; the laws of nature are fixed and initial conditions contingent, but which level of description counts as a 'law' is a modeling choice. If the split is chosen by the user, then computation is doubly subjective: the user decides both the partition and the interpretation. Consequently, any physical process can be made a computation by picking a suitable partition and purpose, and the central claim cannot distinguish computation from non-computation. The paper criticizes Eliasmith's 'alterable rules' as 'not well defined,' but its own 'contingent environment' inherits the same problem without a formal account.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that a physical process is a computation only when it is used to externalize thought, and defines computation as the performance of one or more fixed processes within a contingent environment. It criticizes Eliasmith's 'alterable rules' definition for being ill-defined and instead places the criterion on interpretability: a computational process can always be separated into a fixed part and a contingent, alterable part. The paper reformulates the Church-Turing thesis as a claim about expressibility of rigorous thought as software, and endorses the agent-based model of computation, while conceding that the multi-scalar environments essential to that model have no formal formulation. The argument proceeds through historical examples of thought tools (sundials, abaci, geometric tools, logic machines) and informal philosophical reasoning.","tokens_in":10901,"tokens_out":3689,"duration_ms":36556,"significance":"If the central claim were fully worked out, the paper would offer a genuinely distinctive position on the nature of computation, one that avoids both pan-computationalism and the need for a mind-independent notion of computation. It usefully diagnoses defects in existing definitions and proposes an interpretability-based alternative that connects computation to human thought. The paper is honest about its own gaps, explicitly admitting that the multi-scalar environment is not formally modeled and that the anti-Platonist premise requires further work. Its historical narrative and concrete examples are instructive. The significance is therefore primarily conceptual and programmatic rather than formal; the paper does not deliver a rigorous definition or theorem, but it stakes out a coherent philosophical claim worth further development.","major_comments":[{"comment":"The assertion 'we can always separate a computational process into its fixed part and its contingent or alterable part' is load-bearing for the paper's definition, but no criterion for making this split is given. The examples are perspective-dependent: the CPU is called fixed and memory contingent, yet microcode is alterable; the balance scale is fixed and the objects loaded contingent, yet the scale's construction is itself a contingent outcome of manufacturing; the laws of nature are fixed and initial conditions contingent, yet which description counts as a 'law' is a modeling choice. If the user chooses both the partition and the interpretation, the definition becomes doubly subjective and cannot distinguish computation from non-computation. The paper's criticism of Eliasmith's 'alterable rules' as not well defined applies equally to its own notion of a 'contingent environment.' This needs either a formal criterion or an explicit acknowledgement that the split is entirely stipulated, with the consequences for falsifiability drawn.","section":"§3 (fixed/contingent split)"},{"comment":"The paper's endorsement of agent-based computing as 'the current consensus view of how we interact with the world' rests on the existence of multi-scalar environments, but the same section states: 'We do not know how to model the multi-scalar face that nature presents to us' and 'we have not been able to develop a formal model of such an environment.' The support for multi-scalar entities is deferred to the author's own unpublished work [4]. This is not an independent argument; it is self-support. If the central conclusion depends on an environment that has no formal model and on an unpublished reference, the conclusion must be flagged as speculative rather than established. The authors should either provide the missing formal framework or substantially weaken the claim.","section":"§4.2, Abstract"},{"comment":"The anti-Platonist premise that 'ideas exist only within the realm of mental events' is foundational because it underwrites the claim that a physical process becomes a computation only when used to externalize thought. The author acknowledges in footnote 4 that this position 'requires some care' and that 'these issues can be worked out,' but the paper does not work them out. Without this premise, a tree might be instantiating an idea or a computation without any interpreter, which would directly contradict the central claim. Since the premise is neither argued nor defended, the central thesis rests on an unestablished assumption. The paper should either justify the anti-Platonist stance or present the central claim as conditional on it.","section":"§2 (footnote 4)"},{"comment":"The revised Church-Turing thesis—'to be considered rigorous a thought process must, at least in principle, be expressible as a software'—is asserted rather than argued. The paper distinguishes computational activities from computable functions and claims that the equivalence of models shows they are equivalent as programming languages, but it does not justify why rigorous thought must correspond to software expressibility, nor what 'software' means at this level of generality. This revised thesis is used later to legitimate the agent-based model, so it is load-bearing. The author should provide a supporting argument or clearly label the claim as a stipulative proposal.","section":"§3.1 (Turing Machines vs. Turing computability)"}],"minor_comments":[{"comment":"There is a typo in the sentence 'the environment of an fixed process' which should read 'the environment of a fixed process.' Additionally, the section number '3.1' is used twice, for both 'Non-algorithmic computing' and 'Turing Machines vs. Turing computability'; subsequent sections should be renumbered.","section":"§3, §3.1"},{"comment":"The sentence 'We are not arguing that these activates occur sequentially' contains a typo: 'activates' should be 'activities.'","section":"§2.8"},{"comment":"Footnote 10 ends with 'by the Pauli exclusion principle..' with a double period; one period should be removed.","section":"§2 (footnote 10)"},{"comment":"Reference [4], 'Emergence Explained,' is listed as 'to appear in Complexity' and is cited as the source for multi-scalar entities and energy-driven evolution. Since the paper under review is dated 2006 and the reference appears to be unpublished, the author should provide the publication status or a preprint link, and should not rely on it for load-bearing claims.","section":"References"},{"comment":"The claim that abaci and geometers' tools are 'completely independent of non-human physical processes' is imprecise, since human users are physical agents; the intended contrast is with non-human natural processes, and the wording should be clarified.","section":"§2.4-2.6"}],"recommendation":"major_revision","confidential_remarks":"The paper is an essay-length position statement rather than a technical contribution. The author is candid about the missing formal model, which is commendable, but the central definition relies on an unformalized fixed/contingent split and on an anti-Platonist premise that is acknowledged to require further work. The heavy reliance on unpublished self-citation [4] for the multi-scalar environment should be addressed in revision. I would not recommend rejecting outright, as the paper broaches an important conceptual question with some fresh examples, but the revision needs to either supply criteria for the split or substantially temper the strength of the claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis one is a philosophical essay, not a formal result. The interesting move is the reformulation of the Church-Turing thesis: instead of a claim about which functions are computable, it's a claim about which thought processes can, in principle, be expressed as programs. That is a genuine reframing, and it connects naturally to Wegner and Goldin's interaction machines. The paper also earns credit for the clarity of its examples—tree rings, sundials, balance scales—and for being upfront about what it does not know: it explicitly says no one knows how to model multi-scalar environments, and it concedes that its anti-Platonist stance needs more work. That honesty is rare.\n\nThe soft spot is exactly where the stress-test note lands. The fixed/contingent split is load-bearing and it is stipulated, not argued. “We can always separate a computational process into its fixed part and its contingent or alterable part” is asserted, and every example is a user-relative partition: the CPU is fixed until you reflash the microcode, the balance scale is fixed until you rebuild it, the laws of nature are fixed until you change your model. Since the user chooses both the partition and the interpretation, the definition can make any physical process into a computation. The paper's response—that a process counts only when we use it to externalize thought—does not rescue this, because “use” and “externalized thought” are doing the work without independent criteria. So the claim risks being unfalsifiable. That is not fatal for a conceptual stance, but the paper sometimes writes as if it is a theorem.\n\nThe citation pattern is thin: the key references to multi-scalar entities point back to the author's own unpublished “Emergence Explained,” which is a self-support problem. And the last section on evolution and energy feels bolted on; it does not follow from the definition and could be cut.\n\nWho is this for? Someone thinking about the philosophy of unconventional computing or agent-based models gets a clear, provocative standpoint. It is not a technical contribution. To make it a stronger paper, the author should engage the observer-relative computation literature (Putnam, Searle) and either formalize the fixed/contingent split or explicitly retreat to a weaker claim about what counts as a useful description. With that revision, there is a decent paper here.\n\nRecommendation: send to peer review at a venue that tolerates philosophy of computing, but expect major revision. It deserves a serious referee, not a desk reject.","headline":"A thought-provoking philosophical essay that reframes Church-Turing as a statement about programs, but the load-bearing fixed/contingent split is stipulated and user-relative, making the central claim hard to pin down.","tokens_in":11289,"tokens_out":3375,"would_cite":false,"duration_ms":29040,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that computation is not intrinsic to nature: a physical process such as a tree's shadow counts as computation only when someone uses it to externalize thought, with a fixed process running against a contingent environment…","keywords":["computation","Church-Turing thesis","externalized thought","agent-based computing","interactive computation","multi-scalar environment","unconventional computation","thought tools"],"falsifier":"Show a naturally occurring physical process—for instance, a mineral formation whose structure is isomorphic to the trace of a specific program—that no agent has ever used or intended to use to externalize thought, and have competent observers agree it is computing; under the paper's criterion it is merely a fixed process until used, so a clear positive answer would falsify the claim, while a clear negative answer would support it.","tokens_in":10282,"feed_emoji":"🌳","tokens_out":9952,"duration_ms":87994,"temperature":0.7,"pith_summary":"Most attempts to define computation look for intrinsic properties of physical or symbolic systems. This paper argues instead that a physical process becomes a computation only when we use it as a way to work with externalized thought: a tree counting years is just a tree until someone treats its rings as a calendar, and a running computer is only moving electrons until its activity is understood as performing a mental task. Computation is the playing out of one or more fixed processes against a contingent environment whose contingencies we set and interpret. The paper therefore reformulates the Church-Turing thesis as a claim about the expressibility of thoughts as programs, not about which functions are computable. It concludes that agent-based computing in an open, multi-scalar environment is the current consensus model of interacting with the world, and that a formal model of such environments is missing.","feed_headline":"Tree shadows compute only when we use them to think","feed_subtitle":"Computation is not intrinsic to nature; it is a fixed process used to externalize thought in a contingent environment.","key_machinery":"The load-bearing distinction is the fixed-part/contingent-part split: every computational process can be separated into a fixed process (the mechanism, be it a CPU, the laws of nature, or a balance scale) and its contingent environment (the memory contents, the weights on the pans, the shadows falling on a dial). Computation occurs when an interpreter both controls those contingencies and interprets the result as externalized thought. The paper's companion device is the reformulated Church-Turing thesis, which turns the thesis from a statement about the class of computable functions into a statement about programs: a thought process is rigorous only if it can, in principle, be expressed as software.","core_discovery":"The paper's central claim is that computation is not an intrinsic feature of physical processes. A process in nature may be considered a computation only when we use it to work with externalized thought; otherwise it is just a fixed process, such as a tree growing rings or electrons moving in a computer. What makes a computation is the separation of a process into a fixed part (the laws of nature, a CPU, a program being executed) and a contingent or alterable part (the environment, the loaded program, the input), with the practitioner controlling and interpreting the contingent part. From this the paper draws two corollaries: the Church-Turing thesis should be read as asserting that rigorous thought is expressible as software, and all real-world computation is environmentally driven non-algorithmic computing, so an agent situated in an environment behaves like an oracle machine with nature as the oracle. The paper further argues that a formal model of multi-scalar environments is an unsolved problem and that evolution can be understood as the increasing ability of dynamic entities to extract information from their surroundings.","pith_inferences":["Beyond the paper: if computation is use-dependent, unconventional-computation research is better aimed at building controllable contingencies and interpretation channels than at detecting computational structure already present in nature.","Beyond the paper: the fixed/contingent split is recursive, since what is fixed at one level (a loaded program) becomes contingent at the next (input to that program), suggesting a hierarchy of computational levels rather than one absolute line.","Beyond the paper: the reformulated Church-Turing thesis yields a testable prediction: any reasoning practice that practitioners accept as rigorous but that cannot be expressed as a program would refute it.","Beyond the paper: requiring environments to allow newly arising entities and kinds, as the paper's open-systems view implies, gives a concrete design target for new agent-based simulation frameworks."],"forward_implications":["A tree growing rings, a shadow crossing a dial, and a running computer are each just a fixed process until someone uses them to externalize thought; the same physical event can be a computation in one setting and not in another.","Because the fixed and contingent parts are relative to the interpreter, there is no principled line between program and environment: a loaded program is the environment for a CPU, and data are the environment for a program.","The Church-Turing thesis, restated as program expressibility, implies that a thought process is rigorous only insofar as it can in principle be written as software.","All real-world computation is environmentally driven or non-algorithmic: the focus shifts from the algorithm a process executes to how an environment shapes a process, making an agent in a real environment akin to an oracle machine with nature as the oracle.","Formulating a model of a multi-scalar environment is an open problem; until it is solved, the paper concludes, we cannot formally model how we interact with the world."],"supporting_citations":[{"why":"Frames the motivating question of what it means for a physical, chemical, or biological system to perform information processing.","marker":"[1]"},{"why":"Provides the interaction-based view of computation the paper endorses over closed function evaluation.","marker":"[2]"},{"why":"Supplies the definition of computation the paper modifies, in which rule-governed state transitions have alterable rules.","marker":"[9]"},{"why":"Establishes the origin of agent-based computing, the paradigm the paper claims is the consensus model of interaction.","marker":"[10]"},{"why":"Defines persistent Turing machines, the formal model of open interactive computation the paper interprets through its program-versus-function distinction.","marker":"[11]"},{"why":"Supports the analogy between open systems in thermodynamics and open information-flow agents.","marker":"[12]"},{"why":"Grounds the paper's treatment of multi-scalar environments and dynamic entities that extract energy, supporting the open-problem claim.","marker":"[4]"}],"fun_headline_variants":["If a tree computes, who's reading the shadow?","Tree shadows don't compute; we compute with them","Computation is a choice, not a property of nature","Nature doesn't compute until we use it to think","Computation is what we do with nature, not what nature does"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the paper's anti-Platonist stance that ideas exist only in the mind of someone actually thinking them; if ideas could exist independently of minds, a physical process could be computing even when nobody uses or interprets it, and the definitional claim would collapse.","fun_headline_variants_meta":{"raw":{"variants":["If a tree computes, who's reading the shadow?","Tree shadows don't compute; we compute with them","Computation is a choice, not a property of nature","Nature doesn't compute until we use it to think","Computation is what we do with nature, not what nature does"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2761,"prompt_tokens":814,"completion_tokens":1947,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":430,"completion_tokens_details":{"reasoning_tokens":1866}},"tokens_in":430,"tokens_out":1947,"duration_ms":11567,"temperature":1.0,"reasoning_tokens":1866,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-28T17:48:58.437382+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Show a naturally occurring physical process—for instance, a mineral formation whose structure is isomorphic to the trace of a specific program—that no agent has ever used or intended to use to externalize thought, and have competent observers agree it is computing; under the paper's criterion it is merely a fixed process until used, so a clear positive answer would falsify the claim, while a clear negative answer would support it.","supporting_citations":[{"cited_title":"Preface [to the first ed ition] from the Editorial Board,","cited_arxiv_id":null,"evidence_quote":"Frames the motivating question of what it means for a physical, chemical, or biological system to perform information processing."},{"cited_title":"Interaction, Computa bility, and Church’s Thesis,","cited_arxiv_id":null,"evidence_quote":"Provides the interaction-based view of computation the paper endorses over closed function evaluation."},{"cited_title":"Computation,","cited_arxiv_id":null,"evidence_quote":"Supplies the definition of computation the paper modifies, in which rule-governed state transitions have alterable rules."},{"cited_title":"Behavior and Express iveness of Persistent Turing Machines,","cited_arxiv_id":null,"evidence_quote":"Defines persistent Turing machines, the formal model of open interactive computation the paper interprets through its program-versus-function distinction."},{"cited_title":"The Role of Agent Interact ion in Models of Computing,","cited_arxiv_id":null,"evidence_quote":"Supports the analogy between open systems in thermodynamics and open information-flow agents."},{"cited_title":"Emergence Explained,","cited_arxiv_id":null,"evidence_quote":"Grounds the paper's treatment of multi-scalar environments and dynamic entities that extract energy, supporting the open-problem claim."}],"review_version":1}