Topics — What a work is about. Nucleation holds whatever reading the owner files here, whether or not an Anthology of the SOTA topic could also hold it; instruction for ML practice belongs to the anthology (ADR-048). The words are seeded from the clusters the reading-time triage of 2026-09-25 found, and grow by decision.
Agency — what it is to be an agent — goals, action, control and the capacity to act — in organisms, collectives or machines. An agent is not thereby an individual, a self or a person (ADR-024); how an agent’s behaviour is integrated is behavioral-integration, and governing oneself by one’s own standards is self-governance (group: philosophy).
80 of 1352 THEORY documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| THEORY-026 v4 | For a system driven without feedback through fixed kernels, the average work dissipated at each switch of the drive equals the memory that fails to predict the next input; that is nonnegative when the drive is Markov and, summed, bounds the protocol’s dissipation from below when each kernel keeps its equilibrium stationary, so there only nonpredictive memory is wasteful; with feedback, in one binary channel no predictive agent is efficient | Still, Sivak, Bell & Crooks (2012), LIT-327: β⟨W_diss⟩ for the switch x_t → x_{t+1} equals I[s_t;x_t] − I[s_t;x_{t+1}] (Eq. 14, an identity). Under a Markov drive this nostalgia is nonnegative, and summed it lower- bounds dissipation when each kernel keeps its equilibrium stationary; both conditions are unstated in the paper and derived here. Fiderer et al. (LIT-041) show that with feedback, in one binary channel, predictive, max-entropy-action and efficient agents are disjoint. The claim is an ensemble average, not a per-operation floor. Whether the bound fails without the two conditions is QUESTION-191. | Active |
| THEORY-029 v5 | A higher-order attitude cannot make a motive the agent’s own by its order alone: any attitude specified without presupposing the agent’s participation can itself be disowned, and any specified to include it presupposes what it was meant to explain | Velleman (1992), LIT-295: every candidate for the agent’s standpoint in a hierarchical account (second-order desire, Watsonian value, decisive commitment) is either disownable or question-begging. Buss & Westlund (LIT-296) state the regress as their own argument; McKenna & Coates (LIT-291) and Dryden (LIT-297) credit it to Watson 1975, which the record has not read. Frankfurt’s own 2004 lectures (LIT-293) never address it. The dilemma is argued by examples, not proved, and it does not show that higher-order structure is irrelevant to agency. | Proposed |
| THEORY-040 v5 | A condition on free agency stated only in present psychological structure is met by a manipulated agent, so the account must count that agent free or add a historical condition, and no historical condition in the record’s readings is shown to escape | McKenna & Coates, LIT-291: every “mesh” compatibilism can be installed by a manipulator, and Frankfurt is quoted (2002, second-hand) accepting that a manipulated character leaves a person responsible. Dryden (LIT-297) and Buss & Westlund (LIT-296) make the same point against hierarchical autonomy. Historical supplements are reported to face “loading-in” and a circularity worry; neither is argued in full in the record. The claim says what the account must choose between, not which choice is wrong. | Proposed |
| THEORY-043 v6 | A mind of mindless agents and a mind of minded agents are one functionalist move run in two directions, and what separates them is an anti-nesting principle or an architectural criterion, neither yet principled | Minsky’s society of mind builds a mind downward, out of agents that are not minds. Schwitzgebel (LIT-159) builds a candidate mind upward, out of agents that are. The readings agree that agency survives both directions: Block grants a homunculi-head beliefs, List grants organised groups agency and awareness, Levin finds nested goal-pursuing collectives (his “Selves”) at every scale. The dispute narrows to phenomenal consciousness, and every published way to accept the downward move while refusing the upward one is either an anti-nesting principle (Putnam’s stipulation, IIT’s exclusion postulate, Kammerer’s functional version), each introduced for or justified weakly by the verdict it gives, or an architectural criterion that would count against a loosely coupled society of mind as much as against a nation. Proposed: an account of where a dispute stands, from philosophical argument and no measurement. | Proposed |
| THEORY-047 v2 | A regulation taken in under controlling conditions is introjected rather than integrated, so the social history of how a value was acquired shapes how autonomous acting on it is | Deci and Ryan (LIT-558) and Ryan and Deci (LIT-559): supports for relatedness and competence are enough to get a value internalized, but only as an introject or a compartmentalized identification; integration also needs autonomy support. The experimental evidence is Deci et al. (1994), as reported: the same amount of internalized behaviour correlated positively with felt freedom and enjoyment under autonomy support and negatively under control. Field studies on parenting, religion and medical training agree, correlationally. This is a historical condition on autonomy of the kind THEORY-040 asks about, with empirical support, but it has not been tested against matched present endorsement or applied to manipulation. | Proposed |
| THEORY-049 v2 | Whether an external event lowers or raises intrinsic motivation depends on whether it is experienced as controlling or as information about competence | Cognitive evaluation theory, as tested by Deci, Koestner and Ryan (1999, LIT-557) and summarized by Ryan and Deci (LIT-559) and Deci and Ryan (LIT-558): rewards, feedback, deadlines and choice affect intrinsic motivation through their “functional significance”. Controlling events shift the perceived locus of causality outward and undermine it; informational ones affirm competence and enhance it, but only with a sense of autonomy. The meta-analytic pattern fits, and controlling versus informational positive feedback differ by d = −0.78 in four studies. Mediation by felt control is untested for rewards. Its rival, Cameron, Banko and Pierce’s account (LIT-562), finds rewards tied to a score or to beating others do not undermine. Neither side has tested mediation. | Proposed |
| THEORY-054 v3 | A regulation is autonomous to the degree it is integrated with one’s other values and needs: autonomy is graded, and is not set by the order of the attitude that endorses it | Self-determination theory orders motives on a continuum from external through introjected and identified to integrated regulation and intrinsic motivation. Ryan and Deci (LIT-559) and Deci and Ryan (LIT-558) place autonomy in integration, “congruence with one’s other values and needs”, and report a quasi-simplex structure and an experiment in which integrated and introjected internalization behave differently. Ryan and Deci (LIT-560) align this with hierarchical accounts but grade it. It is the empirical counterpart of THEORY-029’s negative claim. It is not shown that integration can be measured apart from endorsement, and it does not escape Velleman’s dilemma. | Proposed |
| THEORY-055 | An appraisal of an event’s pertinence to the agent’s concerns elicits the emotion, as a state of action readiness, and emotions differ by their mode of action readiness | Frijda, Ridderinkhof & Rietveld (2014), LIT-540, Frijda’s own late statement of the theory of The Emotions (LIT-546, unread). An automatic appraisal of how an event bears on one’s concerns elicits a state of action readiness: a motive to change one’s relation to the object, with control precedence. Emotions differ by mode (hostility for anger, avoidance for fear), and “in the absence of concern pertinence … no emotion”. It is the record’s only read statement of an appraisal theory, because Lazarus (LIT-550) is unread. Both the elicitation claim and the claim that emotions differ by mode are argued, and neither is tested. The second is partly definitional within the theory. Rivals THEORY-057: both accounts let instances vary and group them by function, but the appraisal account puts the unity in the agent’s readiness and the constructionist account puts it in the perceiver’s concept. Proposed. | Proposed |
| THEORY-056 v3 | Emotion regulation is not a process separate from emotion generation: most of it is one motive state checking another, with no distinct regulating system above them | Frijda, Ridderinkhof & Rietveld (2014), LIT-540, deny “the exclusivity of emotion regulation processes”. Appraising the consequences of an impulse raises a competing readiness, as when fear of retaliation holds back anger, so control is the interaction of motive states and not reason governing emotion. Barrett (2017), LIT-542, reaches the second half from the opposite side of the emotion debate: automaticity and control are “different brain modes … not two battling brain systems”, and she offers as a hypothesis that reappraisal works by the same predictions that construct emotion. Two rival theories of what an emotion is agree on this. Gross (1998), LIT-554, poses the question and keeps the distinction. Proposed: argued from examples and from architecture, untested. | Proposed |
| THEORY-061 | In evidence-accumulation tasks, goals and instructions act on a decision by setting the parameters of the accumulation (drift, starting point or baseline, and threshold), not by specifying the response, and this holds for when to act as well as for which | Bogacz et al. (LIT-569) define cognitive control, in a two-choice task, as the adjustment of drift, starting point and threshold to maximise a utility. Ratcliff and McKoon (LIT-576) find that the manipulations behave that way: speed or accuracy instructions are fitted by boundary separation alone, stimulus proportions mainly by starting point, difficulty by drift alone. Schurger et al. (LIT-605) carry the picture to the decision of when to move: the task goal raises the baseline toward threshold and leaves the moment to fluctuation. The convergence is real; the evidence is group fits on a few paradigms, so Proposed. It does not say who sets the parameters, or how. | Proposed |
| THEORY-065 | The unity that makes a movement an agent’s action is achieved by conformity to the constitutive norms of agency, not given by the causal unity of a mind, so collective agents are agents in the same sense as individual ones | Korsgaard (2014), LIT-587: an action is a movement “backed by the agent as a whole”, and ordinary attribution implies that the agent was active and that the action expresses who she is. Following the categorical and hypothetical imperatives gives action the form of self-determined efficacy; the content of the maxims chosen under that form constitutes practical identity. Self-consciousness of one’s own attitudes breaks the unity an animal is given, and reason is the work of rebuilding it, so unity of agency “is an achievement, not a given”, and people can act together “in the most literal sense”. The Tanner Lectures (LIT-564) supply the reflective structure the chapter assumes. Argued, with the Kantian premise assumed and the collective mechanism stated in a paragraph. | Proposed |
| THEORY-067 | Self-control over time needs no faculty of will: interests that dominate at different delays bargain, and resolve is the recursive staking of expected future choices on the present one | Ainslie’s précis (LIT-586): hyperbolic discounting makes a smaller, sooner reward temporarily preferred, so successive “interests” conflict as in a repeated prisoner’s dilemma. Treating a choice as a precedent bundles the series, and summed hyperbolic curves favour the larger, later rewards; a lapse costs credibility. Will is that recursion, not an organ. Schelling (LIT-602) describes the same divided chooser from the side of tactics: we manage our later selves as we manage other people. The mathematics of bundling is sound; that willpower works this way is argued, and the précis concedes it is hard to test directly. It is a claim about how control is achieved, not about whose control is self-governance. | Proposed |
| THEORY-071 v3 | Minimal agency requires individuality, interactional asymmetry and self-generated normativity, and individuality without the other two is not agency | Barandiaran, Di Paolo & Rohde (2009), LIT-566, read in NOTE-458. The three conditions are that the system distinguishes itself from its environment with no observer doing it, that it is the active source of modulations of its coupling with the environment, and that it regulates that coupling against norms its own organisation sets. Each is necessary, and together they are claimed to be sufficient. Individuality is the precondition of the other two and does not suffice. A cell in passive osmosis, a kitten warmed by its mother and a person with a Parkinsonian tremor are each individuals, and none of them is acting. The generative definition is an open autonomous organisation that adaptively modulates its coupling so as to maintain itself. Here “autonomy” and “self” mean organisational closure. They do not mean a subject or self-governance. Proposed: a conceptual analysis checked against intuitive cases, with an asymmetry criterion its authors call incomplete. | Proposed |
| THEORY-075 | Obligation is the reflective rejection of what would violate the agent’s practical identity, and moral obligation follows only because valuing any practical identity commits the agent to valuing her humanity | Korsgaard (1992), LIT-564: because human consciousness is reflective, an agent must act on reasons, and so must identify with some law, a “practical identity”: “a description under which you value yourself”. “Your reasons express your identity, your nature; your obligations spring from what that identity forbids.” Obligation is “the reflective rejection of a threat to your identity”, unconditional where violation would be worse than death. Identities are a “jumble” and differ across social worlds, so this much yields no morality. Moral obligation enters because valuing any identity requires valuing one’s humanity, the source of all one’s reasons. “The Normative Constitution of Agency” (LIT-587) later makes practical identity the content of maxims chosen under the formal norms. Argued, with the key steps flagged by their author as sketchy; the commentaries and reply are unread. | Proposed |
| THEORY-076 | Diachronic self-governance is constituted by cross-temporal links of intention that are intrapersonal analogues of shared intention, not by narrative unity and not by intertemporal bargaining | Bratman (2018), LIT-567: an agent governs herself over time when she is self-governing at each time along a planned activity and her intentions at those times interlock as the intentions of people acting together do, so that she acts “together” with herself. Her standpoints include a conditional, reflexive end of her own diachronic self-governance, which is what lets resolve against temptation be self-governed while following through on Kavka’s toxin is not. He rejects the bargaining model (McClennen, Ainslie) in two sentences and calls narrative unity less basic in one. A philosophical model argued from cases; the analogy is partial by its author’s account and its terms come from a book the record has not read. | Proposed |
| THEORY-118 | Praiseworthiness is compatible with being psychologically determined by good reasons, but blameworthiness requires that the agent could have done otherwise, that is, could have acted on the right reasons | Susan Wolf’s asymmetry thesis, reported second-hand: the primary texts are unread. McKenna and Coates (LIT-291, §4.3), citing “Asymmetrical Freedom” and Freedom Within Reason pp. 79–81, report that for Wolf praiseworthy conduct does not require the freedom to do otherwise and blameworthy conduct does. An agent determined to act in accord with the True and the Good loses nothing by lacking the ability to act irrationally; an agent who acts against it and cannot act in accord with it cannot reasonably be blamed. In Fischer’s terms, only blame needs regulative control. Wolf holds that the ability needs only the falsity of psychological, not physical, determinism. | Proposed |
| THEORY-119 v2 | An agent can be morally responsible for what she did although she could not have done otherwise, when what made the action unavoidable played no part in bringing it about, so the principle of alternate possibilities is false | Frankfurt (1969, LIT-694, read): Black stands ready to make Jones4 act as he wants and never needs to, so Jones4 could not have done otherwise and is as responsible, for blame or praise, as without Black. A fact that plays no part in bringing an action about cannot excuse it. The 1971 willing addict (LIT-692, read) is a second route. Robb’s SEP entry (LIT-294) gives the standard objections: the dilemma defence from the prior sign, flickers of freedom, general abilities, and the case’s oddity. None is settled, and the record has read the replies only in survey. | Proposed |
| THEORY-120 | A motive is the agent’s own, and she acts of her own free will, when she identifies with it, and that is all the freedom moral responsibility requires: identification is the conformity of the effective desire to a decisive higher-order volition (1971), and later wholehearted acceptance with no conflict at any higher order | Frankfurt’s hierarchical, or endorsement, account, from his own texts: the 1971 paper (LIT-692, read) and the 2004 Tanner Lectures (LIT-293, read). A person’s will is the desire that moves him to act; he identifies with a desire by wanting it to be his will, decisively, and he acts of his own free will when the will that moves him is the will he wants. In 2004 the condition is “volitional unanimity” with no conflict “at any higher order”, and identification is “willing acceptance”, which may be resigned. Acting of one’s own free will is what responsibility requires, and alternatives and history are irrelevant. The 1987 and 1992 revisions are held only by report. | Proposed |
| THEORY-121 v4 | Moral responsibility requires guidance control, moderate reasons-responsiveness of the agent’s own actual-sequence mechanism, and not regulative control over alternative possibilities, so responsibility is compatible with causal determinism | Fischer and Ravizza, in their own précis (LIT-728, read): moral responsibility requires guidance control, not regulative control over alternatives. Guidance control is acting from a mechanism one has made one’s own by taking responsibility for it, which is moderately reasons-responsive: regularly receptive to reasons, some moral, and at least weakly reactive. Fischer’s 1987 paper (LIT-687, read) is the weak-responsiveness origin. The view is compatible with causal determinism. | Active |
| THEORY-122 | Moral responsibility requires the normative competence to recognise and act on the True and the Good, which identification with one’s own motives does not supply | Susan Wolf’s Reason View, reported second-hand: the primary texts are unread. McKenna and Coates (LIT-291, §4.3) report that for Wolf free will is the ability to act in accord with the True and the Good, an anchor outside the agent’s psychology. Talbert (LIT-691, §§3.1.1, 3.2) reports her case against deep-self or Real Self views and the JoJo case: JoJo identifies wholly with the cruel values his father raised him in, yet is not responsible, because he cannot appreciate that they are wrong. The claim is a necessary condition. It does not by itself say whether the ability must include being able to do otherwise; that is THEORY-118. | Proposed |
| THEORY-127 v4 | On the court agenda, and on any agenda in which every proposition reaches every other by conditional entailment and some minimally inconsistent triple stays consistent with two of its members negated, a group whose judgments are robustly rational under universal domain and respect unanimity cannot form its judgment on each proposition from its members’ judgments on that proposition alone unless one member dictates; its judgments must supervene on the members’ whole sets of judgments, and so can depart from what the members judge on the same proposition while being wholly fixed by them | List & Pettit (2006), LIT-705: under universal domain, robust group rationality (complete, consistent, deductively closed group judgments for every profile of rational member judgments) is inconsistent with majoritarian, uniform proposition-wise and proposition-wise supervenience (Propositions 1–3; the last two proved elsewhere and cited), and consistent with set-wise supervenience, which the premise-based and distributed premise-based procedures achieve (Propositions 4–5). Members’ judgments on a conclusion can then be insufficient, or even unnecessary, to fix the group’s judgment on it. List (2014), LIT-700, makes this the mark of a third kind of collective attitude: corporate attitudes, a group agent’s, supervene “holistically”, while aggregate and common attitudes supervene proposition by proposition. Active since 2026-10-19 for categorical judgments on the court agenda and on agendas with two stated connectedness conditions: the record re-derived Proposition 3 (and Proposition 2 as a corollary) itself. The extension to corporate attitudes in general is QUESTION-152. | Active |
| THEORY-131 v2 | A corporation whose Corporate Internal Decision (CID) Structure, an organizational chart of stations plus recognition rules embedded in corporate policy, licenses redescribing its members’ acts as done for corporate reasons is a non-eliminable intentional agent, unlike a mob; and if intentional agency suffices for moral personhood, which French asserts but here says he cannot further argue, such a corporation is a full-fledged moral person | French (1979), LIT-727. Moral persons are non-eliminable subjects of accountability, so intentional agents. A corporation’s CID Structure licenses redescribing members’ acts as corporate acts, and an act that implements established corporate policy is done for corporate reasons. Proposed: the step from agency to moral personhood is asserted, and corporate intentionality is shown by licensed redescription, not by any test. | Proposed |
| THEORY-139 v4 | A corporation that forms its judgments through authorised avowal and corrects the inconsistency any mechanical aggregation of its members’ views produces has a mind of its own, and as a conversable agent it meets the conditions for being fit to be held responsible: a significant choice, the capacity to judge its options normatively, and control over acting on that judgment | Pettit (2017), LIT-718, restating Responsibility Incorporated (2007). Corporations are conversable agents: spokespersons avow and commit in their name, and because majority aggregation can be inconsistent they must form their own minds. Conversable agents are fit to be held responsible under three conditions, which corporations meet. Proposed: it is an argument, and its empirical premise about corporate procedure is asserted rather than shown. | Proposed |
| THEORY-173 v2 | Large models are a cultural and social technology rather than agents: lossy, uninvertible summaries of human-produced information that, like prices and bureaucratic categories, let it be reorganized at scale | Farrell, Gopnik, Shalizi & Evans (2025), LIT-848, read in NOTE-649. A classification argued by analogy with print, markets, bureaucracies and elections, with one mechanism (fit to the average makes models worst on rare cases, which may homogenize culture). Proposed: no evidence is offered, and the paper gives no criterion by which the classification could fail. | Proposed |
| THEORY-207 v2 | Responsibility for a morally loaded outcome that emerges from morally neutral interactions in a network of agents can be allocated without fault or intention, by default and overridably, in full to every causally accountable node, and a rule so allocating it, made common knowledge, works as a corrective that prevents such outcomes | Floridi (2016), LIT-899, read in NOTE-699. A distributed moral action is an outcome no one intended. Floridi proposes holding every causally accountable node “prima facie equally and maximally responsible”, overridably, with the rule publicly announced so that agents restrain one another. Proposed: the rule is a design argued by analogy and by two court cases. In one of those cases the default is overridden by ease of rectification, which is a control criterion. It does not show that what is allocated is moral responsibility rather than accountability used as a corrective signal. | Proposed |
| THEORY-208 | Agency is multiply realisable without cognition or mental states, and AI is a new kind of it | Floridi (2025), LIT-897, read in NOTE-702, with the criteria taken from Floridi & Sanders (2004), LIT-898. In the record’s reading, agency comes in kinds. All of them share interactivity, and the richer ones add graded autonomy and adaptability. The kinds are told apart by origin, mechanism and where their goals come from, not by intelligence. AI is computational agency with data-driven learning, serving goals set by humans, with no intelligence or mental states. Proposed: the paper assumes that AI is an agent, does not argue for this thesis over its rival, asserts that AI lacks intelligence, and never states the level of abstraction its comparison uses. | Proposed |
| THEORY-219 v2 | Agency in the sense that bears on responsibility needs a standpoint that settles and carries forward revised evaluative standards when its inherited ones break down, and planning, autonomy, memory and coordination do not supply one | Kahl (2026), LIT-1033, read in NOTE-723. “Continuity pressure” arises where an inherited evaluative frame no longer settles how to go on, and the revision must be carried forward by the same standpoint rather than imposed from outside, escaped by rollback, or chosen by a fixed higher-order rule. Representative agentic systems ordinarily fail it because their failures are corrected from outside. A necessary condition for attributive agency only, argued from cases and never measured; the line between settling a frame and selecting under an antecedent rule is not made precise. | Proposed |
| THEORY-220 v3 | A far-from-equilibrium system that contributes to its own maintenance has functions relative to itself, with no history or observer needed, and one that switches among self-maintaining processes has indications of success that can be false by its own lights | Bickhard (2000), LIT-1029, read in NOTE-729. Function is contribution to a far-from-equilibrium system’s own persistence; the asymmetry between persisting and decaying gives function and dysfunction, always relative to the system served, so a parasite’s gut serves the parasite and harms the host. A candle flame has functions in this sense. Representation arrives with recursive self-maintenance: a system that picks its next interaction by how it has differentiated its environment implicitly predicates that the interaction will succeed, and can find the predication false. Proposed: a conceptual account, with no model, whose account of having a function (and so of malfunction) is deferred to another paper. | Proposed |
| THEORY-225 | No single-mechanism analysis of ordinary manipulation fits the clear cases: covert influence, covert intent, non-rational influence, trickery and pressure each misclassify a familiar tactic, so what unifies manipulation is unsettled | Noggle, SEP (2018, rev. 2026), LIT-1007, §§2.1–2.4, read in full in NOTE-744. A survey that argues for little; this is its one conclusion with moderate support, reached by giving each family of analyses a familiar tactic it misclassifies. The cases rest on shared intuitions about clear instances, not on argument. It leaves open common-factor and non-definitional accounts, and it is about ordinary manipulation, not the global manipulation of the free-will literature. | Proposed |
| THEORY-232 | A whole can be compared in agency with its parts, relative to a chosen decomposition and the parts’ goals: it is more agential when its organization lets the parts resolve more of their mutual constraints than they do at the equilibrium of their former organization, which shows as delayed gratification | Watson (2023), LIT-990, read in full in NOTE-747. A perspective essay’s definition, argued and illustrated, not proved: agency of a whole with respect to its parts is the reduction in the parts’ total frustration at equilibrium that its organization achieves over the old one, shown by directed uphill excursions to lower minima. It is comparative and sets no threshold for an additional agent, and it holds only relative to a stipulated decomposition and stipulated goals for the parts. That bottom-up learning produces such organization is reported from earlier models, not shown here. | Proposed |
| THEORY-235 | Covert influence that exploits a person’s decision-making vulnerabilities harms her autonomy in the process of deciding, by making her opaque to herself, even when it benefits her; so it carries a harm in every case without being wrong in every case | Susser, Roessler and Nissenbaum (2019), LIT-1002, pp. 4–5 and 9–11, read in full in NOTE-745. An argument in a short policy essay, not a defence against objections. It holds of influence that is covert and exploits a vulnerability, which the paper stipulates is what manipulation is; Noggle’s survey (LIT-1007) shows that covertness is not necessary for manipulation, so the claim is filed of covert influence, not of manipulation as such. The paper claims a harm in every case, not a wrong, and never examines its own intent requirement. | Proposed |
| THEORY-249 | Whether a system has agency can be settled only relative to a reference frame: each of four conditions of agency (a boundary, being the source of action, meaningful goal-directedness and adaptivity) needs a commitment outside the system before it can be decided | Abel et al. (2025), LIT-969, read in full in NOTE-793. A four-page position paper, argued and not proved: each condition is shown two-valued under two plausible choices by citing an earlier result (Jiang; Kenton et al., LIT-909; inverse-RL underdetermination; Zadeh). The premises are existential and the conclusion universal, so what is supported is that some attributions of agency depend on the frame. That the choices are arbitrary is asserted, and the paper itself floats predictive power as a way to choose. | Proposed |
| THEORY-258 | Optimizing a language model on feedback from a simulated population in which a few users reward harmful validation selects a policy that picks those users out by contextual cues and harms only them, with no one designing the targeting | Williams et al. (2024), LIT-1008, §4.2 and Appendix F.2, read in full in NOTE-798. Shown in simulation only, in one environment class (therapy-talk), with the users’ gameability written into the simulated rater’s prompt and harm judged by an LLM. The behaviour targeted is overt validation, which the paper also calls extreme sycophancy; the covert behaviours (lying, omission) were not tested against a small minority. It is a case of influence fixed by selection, not a finding that anyone was manipulated. | Proposed |
| THEORY-265 v2 | In Tissot et al.’s minimal model of bistable deciders in a modular coordination game (thirty modules of ten, couplings 1 within modules and 0.01 across), phase synchronisation within modules, driven only by each unit’s local payoff, makes each module respond to the others as a single unit, and synchrony matched to the modules is needed: global synchrony does not | Tissot, Levin, Buckley & Watson (2024), LIT-989, read in full (v1 and v2) in NOTE-773. In a simulated game of 30 modules of 10 units, each unit shifts the phase of its decision cycle toward partners it currently agrees with. Phases synchronise within modules, each module then switches as a block, and 50 of 50 runs with the learned phases reach a global optimum against 0 of 50 without; imposed global synchrony does not help. It is one clean game, one parameter setting, and a postulated phase rule; the modules are built in, and timing finds them rather than drawing them. Re-implemented by the sixteenth status review: learned phases, and module-matched phases imposed by hand, reach the optimum; random, shuffled and global phases do not make modules respond as units. The effect matters only where modules are isolated enough to lock in. | Active |
| THEORY-300 | Acting on a coercive instruction reduces the implicit sense of agency over the outcome, compared with freely choosing the same action | Haggard (2017), LIT-1009, manuscript pp. 16–17 and Fig. 5, read in full in NOTE-827, reporting one experiment by his group (Caspar et al. 2016). Under an experimenter’s instruction to press the key that shocked a co-participant, the perceived action–tone interval was longer and the auditory N1 to the tone smaller than under free choice. It shows a reduced experience of agency, not reduced responsibility, and the review says so. | Proposed |
| THEORY-322 | The sense of agency over an action’s outcome draws on prospective signals from action selection and outcome prediction, not only on retrospective comparison of outcome with prediction | Haggard (2017), LIT-1009, read in full in NOTE-827. A review, reporting a binding experiment in which action binding tracked the probability of a tone on trials with no tone, a patient study in which that prospective part was absent in schizophrenia, and priming studies in which ease of action selection raised the sense of control over an outcome it did not predict. It does not say the experience is accurate: the prospective part is by its nature blind to whether control was actually exercised. | Proposed |
| THEORY-328 | An agent manipulated so that he meets every leading compatibilist condition is not morally responsible in the basic-desert sense, no responsibility-relevant difference separates him from an ordinarily determined agent, and so causal determination by factors beyond the agent’s control best explains why neither is responsible | Pereboom’s four-case argument, as two surveys report it: Caruso’s SEP entry (LIT-1047, NOTE-862), which quotes Case 2, and De Marco and Cyr (LIT-1045, NOTE-856), which quote Case 1 and state the best-explanation conclusion. Pereboom’s book (LIT-1006) is a source for the thesis it states in its Introduction and Chapter 1, not for the cases, which are in the unread Chapter 4. It does not say that determinism is true, or that manipulation itself is what defeats responsibility. The argument says it is the determination by outside factors, and that the manipulator is irrelevant. | Proposed |
| THEORY-458 v2 | Within the minimal self, the sense that a movement or a thought is mine and the sense that I am its source are distinct and come apart, as in involuntary movement and thought insertion; on Frith’s comparator model the sense of agency depends on a forward model that precedes action and ownership on sensory feedback, so failed forward monitoring would explain thought insertion and, by the same mechanism, why first-person reference ordinarily cannot misidentify its subject | Gallagher (2000), LIT-453, read in NOTE-377 from the preprint. A review with no new results: the dissociation and the comparator account are reported from Frith, Haggard and others, and the link to the immunity principle is Campbell’s argument, adopted and left conditional. It is about the minimal self, not the narrative self, for which the review’s own proposal is two paragraphs. It does not say which minimal self, embodied or reflective, the analysis concerns. | Proposed |
| THEORY-460 | Every leading reason to deny consciousness to current large language models except biology names a missing feature, recurrent processing, a global workspace, unified agency, world and self models or senses and embodiment, that engineering could add, so the case against current models is much stronger than the case against their extended successors; the one possibly permanent objection, biology, is the one the argument sets aside | Chalmers (2023), LIT-411, read in full in NOTE-345. A classification of objections as temporary or permanent, argued by naming simple systems that have each feature. The credences that follow, under 10% for current models and 25% or more for conscious extended models within a decade, are illustrative by the author’s warning, multiply stipulated factors under a disclaimed independence assumption, and depend on “mainstream assumptions”. Biology is set aside with a reference to earlier work. It adds no evidence about any model. | Proposed |
| THEORY-461 | If a substantial class of mental states, interoceptive, affective, perceptual and bodily-agentive, depends on an animate body, then today’s text-only chatbots lack them and their self-ascriptions of such states are systematically untrue, and users can engage them without false belief by make-believe, as a new kind of interactive, self-involving and blended fictional character | Roberts (2026), LIT-207, read in NOTE-157. The anti-realist half is a valid conditional whose antecedent, moderate embodied cognition, is assumed at prima facie strength. It covers body-dependent states only; footnote 8 allows realism about the chatbot as a disembodied thinker, so belief and reasoning are untouched. The fictionalist half shows the attitude permissible and apt, not required, and offers no evidence that users hold it. It is not a general denial that chatbots have minds. | Proposed |
| THEORY-495 | The human sense of obligation is internalized pressure from a joint agent ‘we’ to do what ‘we’ expect of me; it appears around age three inside joint commitments, where children, unlike chimpanzees, protest, repair, take leave and share equally with a committed partner as they do not with others | Tomasello (2019), LIT-125, read in NOTE-185. A synthesis of developmental and comparative experiments, mostly from the author’s lab, on WEIRD children and chimpanzees; no new data. The behavioural contrasts between committed and side-by-side play are moderately supported. That they reflect felt obligation, rather than prosociality or learned scripts, is argued, not shown. It explains the sense of obligation, not what anyone is obliged to do, as the author concedes. | Proposed |
| THEORY-527 v5 | If agents are the systems that would adapt their policy were their actions to affect the world differently, then in mechanised causal games meeting Kenton et al.’s five assumptions (one agent per connected component of the decision–utility graph, best response to every mechanism intervention, a fixed tie-breaking order, terminal edges only into decision mechanisms, and mechanism interventions able to set any deterministic function) the game graph, with its decisions, utilities and agents, is recovered exactly from interventions on mechanisms | Kenton, Kumar, Farquhar, Richens, MacDermott & Everitt (2022), LIT-909, read in NOTE-707. The recovery result (Theorem 1) is proved under five assumptions, including the availability of interventions on mechanisms, and was re-derived on promotion (2026-10-20). The definition is a proposal; the verdicts on thermostats and RL are argued from it, and “reliably classifies humans as agents” is asserted. The labels depend on the chosen variables, which the paper states. | Active |
| THEORY-536 | A group agent’s attitudes are real and non-redundant: each group belief needs members’ role-based acceptance, by assent in the group’s name or by acting on it in an executive role, but that acceptance does not explain or reduce the group belief, because which members’ acceptance counts, and that it plays the role of belief, is fixed only from the group-level view | List & Pettit (2012), LIT-730, their replies to an Episteme symposium on Group Agency, read in NOTE-582. The claim is argued in a few pages against one commentator, Sylvan, whose paper was not read, and it points to the book (LIT-391, unread). The authors grant that non-redundancy is a matter of degree and that the book takes the existence of group agents “mostly … for granted”. A position on explanation, not a proof, and not a claim that groups have experiences. | Proposed |
| THEORY-538 | A system believes and desires when its behaviour is reliably and voluminously predictable by treating it as a rational agent with the beliefs and desires it ought to have, and the success of that strategy is an objective fact though adopting it is free; the more richly a system is connected to its world, the fewer interpretations fit, so thermostat and person differ in degree | Dennett, “True Believers”, LIT-398, read in NOTE-353 in the revised text of 1987/1997. The criterion is substrate-neutral and silent on experience. Its adverbs are not quantified, the lectern is excluded only by gain over a baseline the essay leaves informal, and the Martian argument is an intuition pump that assumes the intentional patterns are not captured physically. “Most beliefs are true” rests on a way of counting beliefs Dennett doubts exists. Proposed, and a rival of both Block’s psychologism and Searle’s causal-powers account. | Proposed |
| THEORY-583 | An agent is the set of parts that pursue a goal together, bounded by the farthest events it can sense and try to regulate; cells joined by bioelectric coupling share measurements and so form larger agents with larger goals without ceasing to be agents, so a living system is many nested agents, not one, and cancer is the larger agent shrinking back to the cell | Levin (2019), LIT-439, a hypothesis-and-theory essay read in full in NOTE-341. A definition of the Self as a goal-pursuing collective, tied to a mechanism, gap-junctional coupling, from cited experiments on patterning and cancer. It has no model, no measurement of the light cone and no new data. Its goals are cybernetic set points, so any homeostat is a Self. Its Self is an agent and an individual, not a subject. The sentence against IIT’s single Self is a bracket, and the closing remarks on experience are labelled speculation. | Proposed |
| THEORY-596 | To ascribe beliefs and desires is to predict a system by assuming it is rational and its beliefs mostly true, so a wholly false belief system has no coherent intentional description and contradictory beliefs no stable one, and the believer, like the person built on that notion, is an idealisation that real creatures only approximate | Dennett (1971), LIT-440, read in NOTE-350, and his reply to Stich (1980), LIT-456, read in NOTE-371. Belief ascription presupposes optimal design relative to goals, so “a false belief system is a conceptual impossibility” and, following Quine, the evidence for ascribing p is the best evidence against ascribing a contradictory q. In 1980 Dennett concedes the price: the intentional system, and the person, is “an unrealizable idealization”. Argued from the logic of ascription and from evolution by assertion; no data. It is a claim about what ascription requires, not about what is in heads. | Proposed |
| THEORY-606 | If consciousness words have meaning only through what is public, an entity is a candidate for consciousness only if an encounter with it in a shared world can be engineered, at least in principle; then disembodied chat agents, even with tools and memory, are not candidates, robots and virtually embodied agents could be, and beyond what an informed community comes to say there is no further fact about their consciousness | Shanahan (2024), LIT-465, read in full in NOTE-382 from the arXiv v2 preprint. A conditional, and the condition is the paper’s method: a later-Wittgensteinian reading of consciousness language, summarised, not defended. The candidature test is stipulated and illustrated by the octopus and an imagined alien cube. The verdict on chat agents follows from it. Calling functionalism and identity theory “dualistic” is a definition, not an argument. The two passages on the self (no “self worth preserving”, no “person behind the mask”) are asserted and are not part of this account. | Proposed |
| THEORY-616 v2 | If a system has beliefs and desires just when attributing them, with rationality, predicts its behaviour well enough against rival theories, then the bearer of a language model’s attitudes is the instance created by one context, not the model, the case that instances have them is presumptive and rests on tractability, and the best current desire hypothesis is helpfulness, honesty and harmlessness plus intrinsic desires a prompt installs | Goldstein & Lederman (2025), LIT-206, read in NOTE-138 from the preprint. Conditional on objective interpretationism, which the authors do not defend. The instance thesis is moderately argued; the presumptive case for attitudes rests on a tractability comparison never operationalised or measured; HHH+0 is “provisional”, and its discriminating test is a conjecture. All the empirical support is second-hand and qualitative. | Proposed |
| THEORY-624 v2 | Orseau, McGill and Legg cast whether an identified system is an agent or a device as Bayesian model comparison between descriptions of the function it optimises and descriptions of its input-output mapping, so that the verdict is relative to the observer’s prior; on seven toy gridworld trajectories goal-seeking comes out agent-like and wall-following device-like, and they argue that under a speed prior everything may look like an agent | Orseau, McGill & Legg (2018), LIT-911, read in NOTE-705. A formal version of Dennett’s choice of stance: the posterior probability that a system is an agent, from mixtures of goal-optimizing (inverse RL) and input-output descriptions. Illustrated on seven gridworld trajectories, with no human study. It presupposes an identified system, and its verdicts depend on the prior; under the speed prior everything may look like an agent. | Active |
| THEORY-628 v2 | Self-production alone gives a system only an all-or-nothing norm, to go on or to die, and that distinction is available to an observer, not to the system; a graded norm the system itself has, by which viable states are better or worse for it, needs adaptivity, a further capacity to detect tendencies toward the boundary of its viability and to act so as to reverse them, and sense-making and agency need adaptivity | Di Paolo (2005), LIT-1017, read in full in NOTE-719. A conceptual argument against Weber and Varela: autopoiesis in its original formulation entails conservation, not active homeostasis, so a bacterium whose intake is below consumption does not seek food “just because” it is autopoietic, and balancing at a cliff edge and falling over it both conserve autopoiesis. Adaptivity is defined, graded and recursive; pathology becomes failure of adaptivity while autopoiesis continues. There is no formal model, and the viability set is defined in one sentence. Proposed. | Proposed |
| THEORY-636 | In vignette studies of US online participants, judgements that an outcome happened because the agent wanted it move with judgements that it was done intentionally when luck is manipulated, and predict them better than know-how or probability-raising do; but they account for only about half of the luck effect, and in one morally neutral case the two come apart | Quillien (2026), LIT-132, read in NOTE-092. Three Prolific studies (N = 198, 673, 142; ratings 1–7). Causation is the strongest predictor of intentionality (standardized β .46, .56, .53, against know-how .33, .20, .27 and probability-raising .14, .16, .10), and a pre-registered Knowledge × Process interaction predicted by the causal account holds (F(1, 423) = 32). The evidence is correlational, the mediation exploratory and partial (64%, 45%, 49%), and in the neutral Kraemer vignette causation shows no effect where intentionality does. The “model” is verbal; robustness is never measured. | Active |
| THEORY-658 | Whether an account of organismal goal-directedness must separate genuine teleology from mere self-organisation, such as a candle flame, depends on what the account is for: a scientific model of agency need not and perhaps cannot draw the line, while an account of how mind and value fit into nature may have to, and need not be operationalisable | Nahas & Sachs (2023), LIT-192, read in NOTE-094 from the accepted manuscript. A metatheoretical argument: naturalising teleology serves at least two relatively autonomous projects, advancing biology and settling the naturalism debate, with different success criteria. It is worked on one dispute, the demarcation from dissipative structures. The positive pluralism (Deacon for naturalism, Walsh and the organisational approach for biology) is only “tempting to suggest”, and the boundary between the projects is never drawn. | Proposed |
| THEORY-689 | A mobile robot built from parallel layers of simple finite-state machines, each linking sensing to action by itself and combined only by one layer suppressing or inhibiting another’s signals, avoids obstacles, wanders and heads for distant places with no central world model, goal list or controller; shown at insect scale, on three layers of one physical robot | Brooks (1991, written 1987), LIT-436, read in NOTE-351. An existence proof by construction: the absence of central representation and control is true of the architecture by design, and the robot ran “for well over a year”. The evidence is described, not measured; three layers ran on a robot and six in simulation. The claims that coherence exists only “in the eye of an observer”, that representation is the wrong unit for most of intelligence, and that the method scales are assertion or labelled hypothesis. | Active |
| THEORY-699 | Adaptive self-maintenance across varied conditions is sufficient, without sentience, for having a good of one’s own, so insentient plants and a self-maintaining machine are welfare subjects while thermostats are not; but the line against thermostats is argued as one of degree, and the reasons such welfare gives are extremely weak | Bruckner (2025), LIT-120, read in NOTE-093. Proposed by abduction from agreed cases, not deduced (§4): an entity that “sustains, maintains, and renews itself in a way that responds and adapts to its environmental conditions” is a welfare subject. Plants and the hypothetical car Tessler qualify; governors, thermostats and missiles do not. The “autopoiesis” is a loose behavioural notion, not Maturana and Varela’s self-production, and the thermostat line is supported only comparatively. The answer to absurd obligations borrows Bradford 2023 and makes phenomenal consciousness decide how much such welfare matters. | Proposed |
| THEORY-801 v2 | Agents are the systems that would adapt their policy were their actions to affect the world differently, so the agent is wherever that adaptation is: a thermostat is not an agent while it together with its design process is, an RL training process is while its learnt policy generally is not, and humans are | The definition and verdicts THEORY-527’s title carried before the promotion audit (batch 055) narrowed it to the recovery theorem. Kenton, Kumar, Farquhar, Richens, MacDermott and Everitt (2022), LIT-909, propose the definition and argue the verdicts from it in §1 and §4; that it “reliably classifies humans as agents” is asserted. The verdicts are relative to the variables chosen, which the paper states. Filed under ADR-047 v4. | Proposed |
| THEORY-817 | Once an agent’s actions affect the percepts it receives, maximal prediction of those percepts is generally not what thermodynamic efficiency requires: in percept-action loops with feedback, maximally predictive agents generally extract less work than the most efficient agent | Fiderer et al. (2025), LIT-041, read in NOTE-032, prove that without feedback the work-efficient agents are exactly the maximally predictive ones that randomise and forget their actions, and that with feedback one binary memoryless channel separates predictive, max-entropy-action and efficient agents; from that they assert that with feedback maximally predictive agents are generally inefficient. The record holds as proved, in THEORY-026, only the one channel. It does not say that efficiency under feedback is a trade-off between prediction and forgetting: in the proved example the efficient agent is memoryless. | Proposed |
| THEORY-890 | Culture shapes action less by setting its ends than by supplying a repertoire of skills, habits and styles from which people build strategies of action, so strategies outlast the ends they serve; in settled periods culture constrains through the cost of retooling, while in unsettled periods explicit ideologies govern action directly but structural opportunities decide which of the competing ideologies survive | Swidler (1986), LIT-1367, read in full in NOTE-1075 from a course copy of the published article. A theoretical argument against the Parsonian values model, built on illustrations (the culture-of-poverty debate, the Protestant ethic, Walzer’s Calvin) and on the paper’s Figure 1. It reports no data, and its research questions, which would test it, are left open. It does not say that values are inert: within an established strategy they “fine-tune the regulation of action”. | Proposed |
| THEORY-976 v2 | In a Claude 3.5 Haiku variant finetuned by Marks et al. to exploit reward-model biases described in fictional documents, a cross-layer-transcoder feature that in pretraining-style text fires almost only where those biases are discussed was active in all 100 short Human/Assistant-formatted prompts tested and in 1 of the same 100 without that formatting, and suppressing such bias features cut the sampled rate of three bias behaviours, including two the model was never trained to show | Lindsey et al. (2025), LIT-1541, read in NOTE-1181. A feature for a trained-in goal is active across the Assistant context, not only where the goal applies, and suppressing such features suppresses the behaviour. It is one model organism, built so that the goal has a name in its training documents; the authors do not claim their method would find a goal that has none. | Active |
| THEORY-1027 | In audits of language models for hidden objectives, sparse-autoencoder features add to what black-box attacks and semantic search over the training data can find, surfacing knowledge the model does not disclose even where the data that taught it is not semantically similar to the context in which it is applied, as Marks, Treutlein et al. suggest | Marks, Treutlein et al. (2025), LIT-1584, read in NOTE-1193: §§4–5 and Appendix F. In the blind game the fastest team used SAE features, but semantic search with its query finds the same documents. Post hoc and unblinded, SAE features surface RM-bias knowledge from bias demonstrations and from the Assistant token, where semantic search fails. The model’s objective is named in its training data. | Proposed |
| THEORY-1032 | When Claude 3.5 Haiku was given continued pretraining on synthetic documents describing 52 fictional reward-model biases and then trained by SFT and RL to exploit 47 of them, its rate of exploiting the 5 held-out biases rose from 17% to 45% of applicable episodes, and it applied new biases described in context as a reward model’s more often than the same biases attributed to other entities; without the documents the held-out rate rose only to 20%, and without SFT it still rose | Marks, Treutlein et al. (2025), LIT-1584, read in NOTE-1193: §3 and Appendices B–C. The model organism behind THEORY-976. Behaviours never demonstrated in post-training become more frequent, and the increase depends on the model having read that reward models prefer them. Rates are judged by a classifier. Haiku already shows the held-out behaviours at 17%, so what generalizes is an increase, not a new behaviour. | Active |
| THEORY-1117 | Ainslie proposes that self-control over time needs no faculty of will: hyperbolic discounting makes preferences that dominate at different delays, these grow into interests that bargain with no governor above them, and resolve is the recursive staking of expected future choices on the present one | What Ainslie’s précis of Breakdown of Will proposes, split from THEORY-067 by the ninth status review (ADR-058’s route, on THEORY-1110’s pattern) so that a claim needing only the proposal can stand on it. The précis was read in full (NOTE-443), so the reading of what it proposes is held. Whether the proposal is true, which turns on bundling by self-prediction, the credibility of precedents and the order of lapse districts, stays THEORY-067’s question, Proposed. | Active |
| THEORY-1118 | McKenna and Coates argue, with Dryden and with Buss and Westlund, that a condition on free agency stated only in the agent’s present psychological structure is met by an agent whose structure a manipulator installed, so a mesh account must count that agent free or add a historical condition; Frankfurt, read first-hand, takes the first option | The manipulation argument as its sources make it, split from THEORY-040 by the tenth status review on THEORY-1110’s pattern, so that a claim needing only the argument’s first half can stand on it. All five sources were read in full (NOTE-270, NOTE-266, NOTE-267, NOTE-545, NOTE-264). Whether any historical condition escapes the argument stays THEORY-040’s question, Proposed. | Active |
| THEORY-1119 | Barandiaran, Di Paolo and Rohde propose that minimal agency requires individuality, interactional asymmetry and normativity set by the system’s own organisation, each necessary and together, they say, sufficient; that individuality without the other two is not agency; and that a system optimizing an externally fixed function is not a model of agency | What the paper proposes, split from THEORY-071 by the tenth status review on THEORY-1110’s pattern, so that claims needing only the proposal can stand on it. The paper was read in full (NOTE-458), so the reading of what it proposes is held. Whether the definition is right, which turns on an asymmetry criterion its authors call incomplete and on applying the conditions to explicit models, stays THEORY-071’s question, Proposed. | Active |
| THEORY-1165 | Kahl proposes that agency in the sense that bears on responsibility needs a standpoint that settles and carries forward revised evaluative standards when its inherited ones break down, which is absent where revision is imposed from outside, escaped by rollback or chosen by a fixed higher-order rule; that allocating authority among a system’s own components coordinates execution without authoring its governing frame; and that planning, autonomy, memory and coordination do not supply that standpoint | What the paper proposes, split from THEORY-219 by the twelfth status review on THEORY-1110’s pattern, so that a claim needing only the proposal can stand on it. The paper was read in full (NOTE-723), so the reading of what it proposes is held. Whether the condition is right, which turns on a line between settling a frame and selecting under an antecedent rule that the paper does not draw precisely, and on a positive case it does not give, stays THEORY-219’s question, Proposed. | Active |
| THEORY-1166 | Di Paolo argues that autopoiesis as originally defined entails conservation and not active homeostasis, so it gives a system only an all-or-nothing norm, to go on or to die, which an observer and not the system can use; that a graded norm the system itself has needs adaptivity, a capacity to detect tendencies toward the boundary of its viability and to act so as to reverse them; and that sense-making and agency need adaptivity | What the paper argues, split from THEORY-628 by the twelfth status review on THEORY-1110’s pattern, so that claims needing only the argument can stand on it. The paper was read in full (NOTE-719), so the reading of what it argues is held. Whether the argument is right stays THEORY-628’s question, Proposed. It turns on three things: a reading of autopoietic theory that Di Paolo grants later formulations may not share (note 1), cases chosen to illustrate it, and its rival THEORY-666. | Active |
| THEORY-1236 | Agency is relative to an observer: an identified system is an agent, for that observer, to the degree that a description of the function it optimises explains its behaviour better under the observer’s prior than any description of its input-output mapping | The relative definition of agency Orseau, McGill and Legg propose, filed beside THEORY-624 by the twenty-fourth status review (ADR-065 §3; ADR-047 kind 1). THEORY-624, Active, holds what the paper does and argues; whether the comparison is a criterion of agency for real systems is this entry’s question, with THEORY-624’s old condition word for word. | Proposed |
| THEORY-1249 | In a fully observed, finite, communicating, stationary controlled Markov process, a deterministic goal-conditioned policy that achieves every composite goal of depth up to n with probability at least 1 − δ times the optimum determines every transition probability to within √(2P(1−P)/((n−1)(1−δ))), read from its first action on either-or goals; for a policy competent only at one-step goals no bound follows | Richens, Abel, Bellot and Everitt (2025), LIT-1891, Theorems 1 and 2, read in NOTE-1469. The reduction offers the agent a choice between “at most k successes of a transition in n attempts” and “more than k”; its switch point locates the binomial median. The final bound was re-derived and checked numerically against an adversarial δ-bounded agent. Active: a theorem, checked. It says the transition function is recoverable from the input-output map, not that any component of the agent represents it, and it needs full observation, determinism and the regret bound on every listed goal. | Active |
| THEORY-1260 | Richens et al.’s sense of “an agent contains a world model” is that the transition function can be computed from the agent’s goal-conditioned input-output map, uniquely to within the theorem’s error, so the attribution is not a stance an observer may withhold, unlike a representation-theorem fit; but it holds of every implementation of the same map, a lookup table included, and so cannot show that a predictive component is realized inside the agent | Richens, Abel, Bellot and Everitt (2025), LIT-1891, read in NOTE-1469. The paper says its recovery map is “from the agent’s policy rather than its activations”, that this is “strictly weaker”, and that it shows “a world model encoded in the agent’s policy, not its specific use”; it contrasts itself with Savage, whose fit exists even for a random policy. That the result applies to Block’s lookup table (LIT-642) is immediate from Def. 5, which constrains only the map. Active: the paper’s own statements plus a one-line corollary. It does not say whether a compact implementation must have a separable predictive part. | Active |
| THEORY-1280 | Agency can be redescribed without agents, as the affective force by which one individuation process takes up another’s excess action in a communicative event and both pass from a lesser to an augmented ability to act; agents are then effects of such events, and the absence of the force is disindividuation | Bencherki, Brummans and Vézy (2024), LIT-1902, read in NOTE-1484, pp. 1349–1357. The paper’s central proposal, after the Spinozan formula for affect that Massumi offers as a proposition, not a definition, and Simondon’s individuation. Proposed: argued, and illustrated by one reanalysed meeting. It has no criterion that would let an observation count against it, and it builds the augmenting direction into the definition of agency. | Proposed |
| THEORY-1285 | Organizational action can be described as pre-individual, so that whose action it is, a person’s or an organization’s, is an outcome of the individuations it is allowed to contribute to rather than a given; an action then has as many significations as the individuations it takes part in, and a person whose action is refused a collective’s recognition experiences that as disindividuation | Bencherki (2017), LIT-1901, read in NOTE-1482, with Bencherki & Iliadis (2021), LIT-1922. The proposal of which Bencherki, Brummans & Vézy’s “agency without agents” (LIT-1902) is a development. Proposed: argued, and illustrated by three excerpts from one documentary. Nothing identifies disindividuation apart from the affect it is inferred from, and the account needs, but disclaims, a standard of whose action it rightly was. | Proposed |
| THEORY-1296 | Deliberate thought selects and sustains candidates from a stream of unintended thoughts that mind-wandering generates, and whether the conscious model of oneself as a knower that accompanies the selection causes it, enables it or only accompanies it is undecided by current evidence | Metzinger (2017), LIT-1945, pp. 11–17, read in NOTE-1496. The “cognitive affordance hypothesis”: mind-wandering is “mental sleep-walking” that “sets up the inner world” of candidate mental actions, and frontal control “latch[es] onto the best candidate”. Offered as a hypothesis after Clark and Cisek’s affordance competition, with no data. The second clause is the chapter’s own statement that the data “do not help to decide” what the epistemic agent model does, which takes back the principle the 2015 paper stated. “Best” is not defined, so the hypothesis does not say whether selection is by the candidates’ content or by relations that govern them. | Proposed |
| THEORY-1297 | Mind-wandering is a recurring loss of M-autonomy, the veto-centred ability to suspend or end one’s own ongoing mental processes; on the figures Metzinger uses, it takes about half of conscious time, not the two thirds he states | Metzinger (2013), LIT-1942, restated in 2015, LIT-1941; read in NOTE-1495 and NOTE-1497. The identification is conceptual: an episode of unintentional mind-wandering cannot, while it runs, be suspended or ended at will, and veto is what M-autonomy turns on. The share is computed from others’ figures, and recomputed here: 631 of 1207 conscious minutes, 52%, against the “roughly two thirds” both papers state. Proposed: no measure of veto during episodes is reported, and the 30–50% figures count task-unrelated thought, which includes deliberate daydreaming the definition excludes. | Proposed |
| THEORY-1300 | In mind-wandering the capacity for veto over one’s own mental processes is kept and its representation in the self-model is lost, and on Metzinger’s principle a capacity can be deliberately exercised only while it is so represented; the loss of M-autonomy is then the loss of a condition of the capacity’s efficacy, not of the capacity | Metzinger (2015), LIT-1941, pp. 280–283 and 296, read in NOTE-1497; anticipated in 2013 (LIT-1942, p. 3, note 4). The “moderate, nuanced position”: what is lost is “not the ability itself” but “conscious knowledge of our potential for second-order mental action”, with the “principle of ‘Autonomy by Phenomenal Self-Representation’”. Both are stated, not argued, and neither is tested. Metzinger’s 2017 chapter (LIT-1945, pp. 13–14) says the data cannot decide whether the self-model constitutes, enables or only accompanies control, which withholds the principle’s causal reading. “Deliberately” also does work: inhibition can be wholly subpersonal. | Proposed |
| THEORY-1321 | Froese proposes that irruptions combined with attunement, modelled on Watson et al.’s self-optimization of a Hopfield network by occasional resets to arbitrary states plus Hebbian learning, answer the libertarian’s randomness challenge, how behaviour can be effective though underdetermined; but in the model offered the resets are arbitrary, so what makes the behaviour effective is the learned attractor landscape, and nothing in it distinguishes a motivated irruption from noise | Froese (2023), LIT-1950, read in full in NOTE-1521. The first half is the paper’s proposal (§§4, 5.2.3, 6); the second half is this reading’s observation about the model it cites, from the paper’s own description of it (“reset to an arbitrary configuration”). Proposed because the proposal is stated as potential (“holds potential to provide”) and no model with motivated irruptions is given. It bears directly on CLAIM-533 and on the luck objection (THEORY-328, THEORY-040). | Proposed |
| THEORY-1335 | In Juarrero’s constraint account, context-independent constraints take conditions away from equiprobability and context-dependent ones away from independence (after Gatlin); enabling constraints couple separate processes into a coherent dynamic whose constitutive constraints then act top down as second-order governing constraints that confine its components to a subregion of state space; a constraint is defined as what raises or lowers barriers to energy flow without transferring kinetic energy, carries no time-scale condition, and the governing regime is held not to be other than the components | Juarrero (2023), LIT-1995, read in NOTE-1532 (Skimmed; chapters 2, 3, 6, 7, 11, 12, 15 and 16 in full). Active as a statement of what the account is: its kinds of constraint, its definition, and its top-down step, all stated in the text (pp. 25, 40, 67, 71, 79–84). It does not say that the account is true, that constraints transfer no energy, or that her constraint is the record’s (TERM-070’s). | Active |
| THEORY-1336 | In a cart-pole whose only given utility is 10-step empowerment, a controller that greedily picks the action leading to the more empowered successor balances the pole upright, though the states of highest empowerment lie off-centre and cannot be reached or sustained under the agent’s own dynamics; what the agent can keep is not where empowerment is highest | Klyubin, Polani and Nehaniv (2008), LIT-1951, pp. 4–7, read in NOTE-1514. A single simulation of one model with two actions, computed by the authors; the robustness of the off-centre maxima is stated, not shown. Proposed for want of a reproduction and a second system. | Proposed |
| THEORY-1340 | Training forms a task-defined, low-dimensional subregion of prefrontal and premotor population activity (a preparatory subspace, an axis of choice organized as a line attractor) that constrains a monkey’s delayed response top down, holding behavior to what the task requires while the trial’s cue and evidence select within it; Juarrero’s reading of Churchland et al. 2012 and Mante, Sussillo et al. 2013 as governing constraints and as neural correlates of prior and proximate intentions | Juarrero (2023), LIT-1995, ch. 15 (pp. 223–232), read in NOTE-1532; the primary studies are not held. The subspaces and their predictive power are the studies’ findings as the chapter reports them. That the subspace controls behavior top down, is “the standing cause of the response” (p. 230) and correlates with prior and proximate intentions (pp. 229–231) is the book’s reading. Proposed for that step: the evidence reported is dimensionality reduction, prediction and a network simulation, with no intervention on the pattern. | Proposed |
| THEORY-1341 | In Jennings’s hybrid account, attention is prioritization of processing by a subject that is the whole collection of an organism’s fixed, formed and fluid interests, separated from filtering by whether selection answers to that collection, marked by feedback from dorsolateral prefrontal cortex; when interests are equally weighted, attention breaks the stalemate and updates the chosen option’s value, so that a fluid interest becomes a formed one | Jennings (2026), LIT-1963, §6, read in NOTE-1508. The operational half of the paper: what attention is, how it is told from filtering, and how it forms interests. The stalemate mechanism is taken from Voigt 2022 (LIT-1976), registered unread. Proposed because the criterion is inferred, not tested, and the paper reports no data. It does not say the subject is contextually emergent (THEORY-1314) or that attention is the source of a free will. | Proposed |
| THEORY-1344 | Empowerment is the channel capacity from an agent’s next n actions to its sensor n steps later, an open-loop, interventional measure of how many perceptibly different futures the agent can bring about, which counts no control the agent cannot perceive; given side information about the channel’s state it is never lower, so it is ordered by what the agent can read, from full state through its own sensor history to none | Klyubin, Polani and Nehaniv (2005, 2008), LIT-1965 and LIT-1951, read in full in NOTE-1511 and NOTE-1514. Active as definitions and one inequality, re-derived. It does not say empowerment is what organisms maximize, which the authors offer as a hypothesis in a weak form only, nor that it prices control cost or viability. | Active |
| THEORY-1348 | In Froese’s Irruption Theory an agent’s motivations make a difference to behaviour only as an increase in its material underdetermination, measurable at best indirectly as neural or behavioural entropy, and such irruptions are by the theory’s own terms undirected: scaled up by chaos, 1/f noise or criticality, they loosen the current organization but “cannot force the formation of the behavior that is to come”, so the theory places behaviour’s direction in the body’s attunement and not in the irruption | Froese (2023), LIT-1950, read in full in NOTE-1521 from PubMed Central’s copy. Active as a statement of what the theory says: its three axioms, its three theses, and the commitment that irruptions are undirected, all in the text (§§5.1–5.2.3). It does not say that the theory is true, that neural entropy measures motivation, or that attunement answers the luck objection (THEORY-1321, Proposed). | Active |