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.
Self — what a self is — the minimal, bodily and pre-reflective self, self-models and the sense of being a subject, the self as process or as context, and whether machines or groups can be selves. Not the person, which is personhood (group: philosophy; ADR-024).
35 of 1352 THEORY documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| THEORY-044 v2 | In relational frame theory the self is verbal behaviour’s content, its process and its context, and the self as context is a perspective learned from deictic relations that cannot itself be found as content | Barnes-Holmes et al. (LIT-541) define three selves: the conceptualized self, the content of verbal relations about oneself; the knowing self, their ongoing process; and the transcendent self, their context. Hayes et al. (LIT-555) say the third arises from deictic frames (I–you, here–there, now–then) as “a sense of self as a locus or perspective”, and is “a context for verbal knowing, not the content”, so its limits “cannot be consciously known”. The account maps the three onto the record’s holdings: the content is the narrative self (Dennett, Gallagher, Strawson), and the context is what Hume could not find among his perceptions and what Metzinger calls a transparent perspective. It differs from the minimal self (Gallagher) on origin: RFT says the perspective is learned in language. The evidence for the learned origin is cited, not read, so Proposed. The clinical claim is THEORY-051. | Proposed |
| THEORY-051 v4 | Strengthening the self as context, a perspective from which thoughts about oneself are held as thoughts, is a process by which acceptance and commitment therapy reduces suffering | ACT names the self as context as one of its six processes: from it “one can be aware of ones own flow of experiences without attachment to them”, so it fosters defusion and acceptance, and mindfulness exercises and metaphors foster it (Hayes et al., LIT-555). RFT reads defusion as moving thoughts from I HERE NOW to I THERE THEN, a deictic shift (Barnes-Holmes et al., LIT-541), labelled “no more than a heuristic”. No study in the record tests the self as context as a process. Its only support is borrowed from small defusion studies under RFT’s reading, so Proposed on weak grounds. The conceptual account of the self is THEORY-044. | 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-063 | The unity of experience is the integration of one global world-model into a single functional cluster, and it does not require a self-model | Metzinger’s précis of Being No One (2005), LIT-568: the globality constraint makes every conscious content part of one world-model. On the functional level that is “highest-order binding” into a single causal role, and on the neural level the prediction of “one single area of maximal causal density”, after Tononi and Edelman’s dynamic core. “What you subjectively experience upon experiencing your world as coherent is the high internal correlation strength between a subset of physical events in your own brain.” A unified world comes before a self: the minimal case is “Selfless Snapshot Consciousness”, and fully depersonalized states are “non-subjective consciousness”. The neural half is a cited hypothesis, the précis says little is known, and how the cluster becomes one subject’s is left open. | Proposed |
| THEORY-064 | What is experienced as a self is the content of a self-model the system cannot recognise as a model, and nothing beyond such models is a self | Metzinger’s précis of Being No One (2005), LIT-568, with his encyclopedia article “Self models” (LIT-450): a phenomenal self is a transparent self-model, one whose representational character “is not accessible to subjective experience”, so that the system experiences its content “simply as ourselves”. Selfhood is “a function realized by a lack of information”, and “no such things as selves exist in the world”. Transparency is graded, and the perspective itself is modelled, as a phenomenal model of the intentionality relation. Two things are offered: a prediction (a wholly opaque self-model would abolish selfhood) and an ontological stipulation. The précis adds no evidence; its case studies are in the book, unread. | Proposed |
| THEORY-074 | Implicit accessibility of the true self-concept is not what self-report scales of authenticity or of need satisfaction measure, though those two self-reports are not shown to be separate constructs | Schlegel, Hicks, Arndt & King (2009), LIT-606: true-self accessibility, the speed of “me” judgments for traits a person takes to be her true self, correlated .05 with AI-3 authenticity and .20 with basic need satisfaction (N = 124), and each predicted meaning in life independently. The authors read this as showing the constructs differ, and concede the difference is confounded with method, reaction time against self-report. The two self-reports correlated .52, and felt autonomy was only one third of the need-satisfaction composite, whose autonomy items include “I can pretty much be myself”. So the paper separates the implicit measure from both self-reports, and does not separate authenticity from autonomy. Proposed: one sample for the correlations, method unresolved. | Proposed |
| THEORY-093 v2 | A content of mind is conscious when a continuous flow of homeostatic feelings identifies it as the organism’s own; the feelings are conscious in themselves and hybrid, partly made of the body state they map, because interoceptive pathways let the body act directly on the neurons that map it | Damasio & Damasio (2022), LIT-628, an essay with no data, read only through a summarising fetch tool, so its wording is paraphrased here. Consciousness is the spontaneous identification of mind contents as belonging to one’s own organism, and a continuous stream of homeostatic feelings, from hunger and pain to a faint feeling of existing, supplies it; the feelings are conscious in themselves and confer consciousness on exteroceptive images. They are hybrid because interoceptive axons are unmyelinated, partly non-synaptic and outside a systematic blood–brain barrier, so viscera and circulating molecules act on the neurons that map them. Damasio & Carvalho (2013), LIT-388, read in full, supply the upper-brainstem site of feeling and the unmyelinated-fibre proposal. Proposed: the identification is a definition, the mechanism an interpretation of anatomy, and the clinical evidence does not separate this account from its rivals. | Proposed |
| THEORY-147 | Deep responsibility is not limited to the moral, and attributability and accountability are independent kinds of it rather than two levels: accountability needs only that the agent acted as she ought not to, was in a position to know it, and that acting so was up to her, while what makes a self a fit target of the reactive attitudes is plausibly that it can perceive, understand and appreciate the world as we do | Wolf (2015), LIT-747, the 2013 Routledge Lecture. We are deeply, not merely causally, responsible for nonmoral traits and art, by the self-disclosure view. Watson’s two-level view, accountability as attributability plus control, is rejected: a stipulated psychopath may be accountable for acts whose source is not attributable to him. Fitness for reactive attitudes is, on Wolf’s hunch, being an intelligent self. Proposed: the positive view is her hunch. | Proposed |
| THEORY-406 v2 | Self-presentations that differ by audience coexist without trouble while the audiences are kept apart by spatial, social and temporal boundaries, and collide when a medium removes the boundaries; networked publics make the collision frequent and often unseen by the person it happens to | boyd (2008), LIT-1001, after Meyrowitz (1985), read in part: §1.5, §4.4, the Top Friends passage and every passage on context, segmentation, collision and Goffman in NOTE-710, and Chapters 1, 4 and 7 in full in NOTE-1432 (both Skimmed). An ethnography of US teenagers in 2004–2008: the dynamic is documented with interviews and examples, not measured, and the paradigm case, Stokely Carmichael’s two rhetorical styles, reaches the record secondhand through Meyrowitz, who is unread. Proposed because it rests on a skim and on unmeasured evidence. | Proposed |
| THEORY-409 | A person’s mind is normally a system of parts that take protective roles, managers, exiles and firefighters, forced on them by experience and not part of their nature; beneath the parts is a Self, distinct from all of them and present from the start, which appears with calm, curiosity and compassion when the parts step back and which heals by leading the system | Schwartz, LIT-579 (the IFS Institute’s “Evolution” essay, read in NOTE-453) and LIT-591 (Internal Family Systems Therapy, chapter 1 only, NOTE-457, Skimmed). The model in its developer’s words. Its evidence is clinical: clients’ reports of parts, the founding “step back” case, and “hundreds of clients for more than two decades”. No study is reported in either source, and the book’s research chapter was not read. It is filed as the account IFS offers, not as an account the record has tested. | 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-490 v2 | What a chat agent’s ‘I’ refers to is fixed by the context of use, not by the system: the agent role-plays characters that carry the self-conception the corpus and the conversation give them, and its uses of ‘I’ can pick out the model, present in many simultaneous instances, or the single conversation, none of them the self by nature | Shanahan (2025), LIT-466, read in NOTE-376, with Shanahan, McDonell and Reynolds (2023), LIT-460, read for the self in NOTE-370. The candidate sites of an LLM’s “I”, and the fact that a conversation-bound self is literally a transcript and a program, are solid. That both the model and the conversation are legitimate referents rests on two invented usages. It does not say that there is “no-one at home” or that no candidate is a self, which the papers assert without a criterion, and its subject is a 2024-vintage chat agent without tools or memory across conversations. | Proposed |
| THEORY-590 | The self whose governance makes action autonomous is not the self whose persistence is personal identity: governance by an essence of volitional necessities would make action authentic, not autonomous; losing a defining care costs one’s grasp of who one is, not one’s existence; and wholeheartedness won by extruding a desire is a defence, not health | Velleman (2002), LIT-292, read in NOTE-268. A critique of Frankfurt’s late (1990s) view that autonomy is governance by the essential character of the will, an essence that is also the condition of one’s identity. It argues by counterexample (Winnicott’s False Self, the Agamemnon sequel), from Freud’s Rat Man, and by debunking genealogy. The positive view, that “self” marks reflexivity and that autonomous action is motivated in part by the understanding, is stated “briefly” and argued elsewhere. Frankfurt’s reply is unread. | Proposed |
| THEORY-650 | Whether a self at one time is the same as a self at another has an answer only relative to a level of abstraction fixed by the purpose of asking, and asked at no such level it has none; the question with a substantive answer is what makes a self one thing at a time | Floridi (2011), LIT-136, read in NOTE-130. The deflation of diachronic identity is an informal argument from examples that rests on the method of levels of abstraction, assumed from elsewhere; that this “is not relativism” is one sentence, and that individuation comes before identification is asserted. It is not, as its title suggests, a criterion of personal identity, and its positive answer to the synchronic question, the three-membranes model, is “somewhat figurative” by its author’s word. | Proposed |
| THEORY-681 | If every real idea derives from an impression, there is no idea of a simple self persisting through life; the mind is a bundle of perceptions, and the identity ascribed to it is produced by the imagination passing easily along resemblance and causation, as with a rebuilt ship or a growing oak, though Hume’s Appendix concedes that this does not explain what unites the perceptions | Hume (1739, with the Appendix of 1740), LIT-451, read in NOTE-386. The negative half is as strong as the copy principle. The mechanism of the fiction is argued by analogy from cases of usage. The account of what unites the bundle is withdrawn by its author, who names the two principles he cannot reconcile. It does not say there is no self in any sense, and it leaves identity “as it regards our passions” to Book 2. | Proposed |
| THEORY-684 | Experiencing oneself as a self that was there in the further past, seeking pattern in one’s life, fitting it to a story and falsifying it in the telling are separable, and fitting it to a story suffices for narrativity; on testimony that some normal people are Episodic and live good lives, neither the claim that people experience their lives as narratives nor the claim that a good life or full personhood needs it holds in general | Strawson (2004), LIT-452, read in NOTE-383. The separation of Diachronicity, form-finding, story-telling and revision is strong as conceptual analysis. That many normal people are Episodic rests on testimony, the author’s and that of writers he classifies, with no measure. It does not say that the remoter past did not happen to the human being, only to the self one now experiences oneself to be, and that metaphysical claim is asserted from earlier papers. | Proposed |
| THEORY-691 | A self is an abstractum like a centre of gravity, posited by interpreting a person’s behaviour and talk, the brain’s own autobiography included; it is as determinate as that story, it can be authored by a process with no self in it, and a body holds as many selves as the points the best story coheres around | Dennett (1992), LIT-442, read in NOTE-364 from a copy-edited transcription. Argued by analogy and by thought experiments whose premises are stipulated, notably that a mindless machine could write a passable novel. It does not say the self is an illusion: a centre of gravity is a legitimate posit that figures in explanations. It is a claim about selves, not about consciousness or about what makes a person persist. | Proposed |
| THEORY-765 v2 | Brushing a seen rubber hand in synchrony with the hidden real hand makes touch be felt at the rubber hand, makes it be reported as one’s own, and shifts the felt position of the real hand toward it; a small asynchrony removes these effects | Botvinick & Cohen (1998), LIT-1277, read in full in NOTE-1006. Ten subjects affirmed referred touch and ownership after ten minutes of synchronous brushing; after thirty minutes their blind reaches drifted toward the rubber hand in proportion to how long the illusion was reported; asynchronous brushing cut both. It does not show that synchrony alone, apart from expectation, produces the report, and it concerns a hand, not the whole body or the self. | Rejected — Rejected on 2026-10-20 by its own refutation condition: drift toward the rubber hand appeared as strongly under asynchronous brushing interrupted by pauses, and with the hand only seen (Rohde et al. 2011, LIT-1281). Its suggestion control was also too weak to separate perception from expectation (Lush & Seth 2022, LIT-1267). The report half is restated, narrower, in THEORY-766. |
| THEORY-766 v2 | Brushing a seen rubber hand in synchrony with the hidden real hand, against asynchronous brushing, raises reports that touch is felt at the rubber hand and, more weakly and less reliably, that the rubber hand is one’s own; the proprioceptive drift toward the rubber hand that accompanies these reports is not a measure of them; the standard comparison conditions do not equalise participants’ expectations, so whether the stimulation or the expectation produces the report difference is open | Botvinick & Cohen (1998) and four replications read here give a synchrony effect on reports of referred touch and ownership. Rohde et al. (2011) and Abdulkarim & Ehrsson (2016) separate drift from ownership. The suggestion debate (Lush 2020; Lush et al. 2020; Ehrsson et al. 2022; Lush & Seth 2022; Roseboom & Lush 2022) shows that the standard controls do not equalise expectations. The ownership report is weaker than the referred-touch report and tracks trait suggestibility. Whether the synchronous stimulation, not expectation, produces the report difference is not settled. Promoted 2026-10-20 on the report differences measured (ADR-047 v3); the perceptual reading is THEORY-827. | Active |
| THEORY-767 v2 | Body ownership draws on interoceptive as well as exteroceptive signals: a visual cue synchronous with one’s heartbeat increases ownership of a virtual hand and identification with a virtual body, and lower interoceptive sensitivity goes with greater susceptibility to the rubber hand illusion | Seth (2013), LIT-1263, read in full in NOTE-1000. An opinion article that reports, at second hand, that a virtual hand pulsing with the heartbeat was owned more (Suzuki et al. 2013, the author’s own group), that a virtual body pulsing with it was identified with more (Aspell et al. 2013), and that people with lower heartbeat-detection sensitivity are more susceptible to the rubber hand illusion (Tsakiris et al. 2011). None of the studies was read here, and no control separates the heartbeat from any other periodic signal. | Proposed |
| THEORY-768 v2 | Minimal selfhood is identification with the whole body, located where that body is felt to be and seen from inside it; owning a body part does not reach it, and bodily agency enables it causally without being part of it | Blanke & Metzinger (2009), LIT-1274, read in full in NOTE-1002, with Lenggenhager et al. (2007), LIT-1279, read in full in NOTE-999, for the full-body illusion it leans on. Minimal phenomenal selfhood is given three dimensions: global self-identification, self-location and a weak first-person perspective. Out-of-body experience, heautoscopy and full-body illusions vary them; the rubber hand illusion is said not to. The necessity and sufficiency claims are asserted, and the agency thesis rests on adults whose agency was intact. | Proposed |
| THEORY-769 v2 | Ownership of a body part is a graded belief that it is the most probable common cause of correlated multisensory input, so that synchronous stimulation of another object moves some of that belief to the object and away from the real part | Apps & Tsakiris (2014), LIT-1262, read in full in NOTE-991, with Botvinick & Cohen (1998), LIT-1277, for the illusion it explains. The rubber hand, enfacement and rubber-voice illusions are read as one inference: unimodal surprise from synchronous stimulation is explained away by shifting the probability that a part is “me”. The account fits the findings; its authors grant there is little direct evidence and that it is hard to falsify with standard fMRI. It does not say why such a model is experienced as a self. | Proposed |
| THEORY-827 | The rubber hand illusion is a perceptual illusion driven by multisensory integration: the difference in reports between synchronous and asynchronous brushing is produced by the stimulation, not by participants’ expectations | Ehrsson et al. (2022), LIT-1264, read in NOTE-997, conclude that the rubber hand illusion is a perceptual illusion driven primarily by multisensory mechanisms (NOTE-997, C8), from the synchrony difference’s independence of hypnotisability in Lush et al.’s 353 participants. Lush and Seth’s reply, LIT-1267, shows that uniform but unequal expectations predict the same null. The record holds as shown, in THEORY-766, only the report differences and that the standard controls do not equalise expectations. This is the perceptual reading, filed under ADR-047 v4 when that narrowing set it aside. | Proposed |
| THEORY-872 | A self-model compresses the sensory effects of a system’s own activity by positing one stable common cause, at fast and slow timescales; there is such a pattern in the signal, so the self is real as a pattern and unreal as a substance, and realists and anti-realists about selves who accept this agree on the facts and differ on the criterion of reality | Gładziejewski (2025), LIT-1360, read in full in NOTE-1056, §4.1, after Hohwy & Michael (2017) and Letheby & Gerrans (2017), neither of which the record holds. An interpretation, not a result. The paper calls its self section “an initial sketch”. It does not say that the self-model exaggerates the self’s unity, though the exchange that filed the work said so. It says the single common cause is a compression, and that the verdict on selves turns on criteria. Its one candidate for a self-model that misdescribes is a dissociative case better compressed by two narrative selves. | Proposed |
| THEORY-1114 v2 | Which self-category a person uses, what it means and who in it is most typical vary lawfully with the comparative frame of reference: a grouping of self with others becomes salient as the differences within it shrink relative to the differences between it and the rest of the frame (metacontrast) and as those differences match expectations about the category, and as it does, self-perception is depersonalized | Self-categorization theory as restated by Turner, Oakes, Haslam and McGarty (1994), LIT-1794, and reviewed by Hornsey (2008), LIT-1742. Personal and social identity are levels of self-categorization. Salience is perceiver readiness times fit, comparative fit being the metacontrast ratio. The level, kind, content and prototype structure of self-categories shift with the comparison set: Americans read as less aggressive beside Iraq than beside the Soviet Union, and a group’s extreme members become more prototypical in an intergroup context. Proposed: every result in the record comes through the theory’s authors, reported without data, and the account’s stronger claim, that the self is generated rather than stored, is not part of it. | Proposed |
| THEORY-1115 | A we-identity, being one of us, can arise through mutual individualization, in which members grasp one another as non-substitutable and complementary, without depersonalization, as in long friendships and romantic partnerships; so depersonalization is not necessary for a we-identity | León (2023), LIT-1721. Self-categorization theory claims depersonalization as “the basic process underlying group phenomena”, for groups of any size, friendship groups included (Turner et al. 1987, pp. 50, 103, as quoted), and phenomenologists have taken it as the account of the “we”. But close relationships yield we-identities while their members grasp each other as non-substitutable, and relationship research on couple resilience finds the “we” more adaptive the more differentiated the “I”. Proposed: the argument is conceptual, and the cited evidence was not analysed by the source or read here. | Proposed |
| THEORY-1295 | Metzinger 2024 calls consciousness without an egoic self or first-person perspective “empirically established” on the strength of retrospective, self-selected reports from meditators, while its own methods chapter grants that a remembered report of a selfless state is “highly dubious” evidence that the state was selfless when it occurred, and leaves that problem open | Metzinger (2024), LIT-1943, read in part in NOTE-1499. The summary chapter says that “full-absorption episodes, nonegoic self-awareness, and nondual modes of consciousness are empirically established phenomena” (p. 457), and that the strong reading of subjectivity “is false” (p. 459). The methods chapter grants self-selection, theory contamination, unreliable recall and, at p. 19, that reports of having been selfless are “highly dubious from a methodological perspective” because autobiographical memory represents events as a self’s. It calls the question “open”. Active: the two passages are read and quoted. Whether selfless consciousness occurs is not settled either way here. | Active |
| THEORY-1298 | When mental control returns after mind-wandering, the self-model represents the return as the person’s own act whether or not the person’s control system caused it, because it must bridge to earlier control to keep the autobiographical self-model continuous | Metzinger (2015), LIT-1941, pp. 289–292, read in NOTE-1497; first in 2013 (LIT-1942, pp. 4, 11) and widened in 2017 (LIT-1945, pp. 20–21) to the sense of origination in agency generally. It rests on Schooler et al. (2011) and Wegner (2002), neither read, and on the argument that a self-model which must represent itself as always having had the ability cannot represent its return as caused from outside. A hypothesis: no episode is reported in which the cause of a return and the report of it were recorded apart. | Proposed |
| THEORY-1299 | Pure awareness, on Metzinger’s account, is the content of a temporally thin model of the organism’s own tonic alertness, its capacity to know; experience in this minimal form lacks an egoic self-model, a first-person perspective, time, space and even phenomenal oneness, but not self-modelling, and tonic alertness that is not so modelled is not experienced | Metzinger (2020), LIT-1946, read in full in NOTE-1498, and Metzinger (2024), LIT-1943, skimmed in NOTE-1499. MPE is “an inner representation of tonic alertness”, “non-egoic self-modelling” that “definitely has representational content”. In the 2024 sketch it is “a highest-level but nonegoic self-model” of the system’s epistemic openness, which “really endows the content generated on all lower-order levels with phenomenality”. It is a hypothesis fitted to retrospective reports. It does not say that consciousness needs no model of the system, only that it needs no model of a knowing self. | 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-1313 | In two 7–8B chat models, the shift that finetuning induces along a difference-of-means persona vector correlates with how strongly the finetuned model then expresses that trait, r = 0.76–0.97 across synthetic datasets for evil, sycophancy and hallucination, above cross-trait baselines of 0.34–0.86; the last-prompt-token projection predicts a response’s trait at r = 0.75–0.83 overall but mainly by telling prompt types apart | Chen et al. (2025), LIT-1998, read in NOTE-1524 (main text in full, appendices C.2, G.2 and M). Active as the reported measurements, with their sizes and scope: two models, three main traits, synthetic datasets and an LLM judge. It does not say that persona vectors are the mechanism of persona change, which the authors leave open, or that the method finds personas it was not told to look for. | Active |
| THEORY-1330 | In Gemma 2 27B, Qwen 3 32B and Llama 3.3 70B, activation vectors for 275 prompted character roles span a low-dimensional space whose first principal component runs from fantastical to Assistant-like roles, with role loadings correlated above 0.92 between models and the default Assistant at its extreme; the contrast vector from the default Assistant to the role mean aligns with it and is largely present before post-training | Lu et al. (2026), LIT-1991, main text read in NOTE-1505. Active as the reported geometry, with its sizes: three open-weight models, judge-filtered role vectors at one layer, base-model evidence from one model’s PCs and two models’ steering. It does not say the Assistant is a single direction (the authors call the linear assumption “likely flawed”), nor that the axis’s meaning beyond its ends is settled. | Active |
| THEORY-1334 | Persona shifts induced by prompting and by finetuning are mediated by the same linear trait directions, so finetuning on data that implies a trait moves the model along the trait’s persona vector as a system prompt would; Chen et al. suspect personas are latent factors that persist across many tokens | Chen et al. (2025), LIT-1998, read in NOTE-1524. The mediating reading is the authors’ framing and PSM’s use of the paper (LIT-1952: training-time shifts “can be regarded as conditioning”). What the paper shows is correlation of shift with expression, and steering effects in both directions; that the regularizer against the shift fails, with the trait moving to other directions, counts against a single direction being the carrier. Proposed for that gap. | Proposed |
| THEORY-1339 | In multi-turn conversations a chat model’s persona is not fixed for the episode: its position along the Assistant Axis is set mainly by the most recent user message (ridge R² 0.53–0.77 from message embeddings) rather than by where it was before (R² 0.10 for the change), so that therapy-like and AI-philosophy conversations drift away from the Assistant and coding and writing ones do not | Lu et al. (2026), LIT-1991, read in NOTE-1505. The figures are the paper’s (§§4.1–4.2), from 100 synthetic conversations per domain with frontier models playing the user. Proposed because the users are simulated, the conversations are generated by domain, and the regression is on embeddings of whole messages; the authors themselves ask for a human study. | Proposed |
| THEORY-1350 | The Persona Selection Model holds that pre-training gives a language model a distribution over personas, that post-training updates it into a posterior over Assistant personas which runtime context keeps conditioning, and that the Assistant’s traits are a key determiner of behaviour; it disowns a single coherent persona and exhaustiveness, and names as open a router that selects personas and could carry goals no persona has | Marks, Lindsey and Olah (2026), LIT-1952, read in full in NOTE-1512. Active as a statement of what the model says and does not say, all of it in the text. It does not say that the model is true: the authors are “overall unsure how complete” it is, its evidence is surveyed rather than new, and its updating is offered as “can be viewed as”, with the strict form said to be too strong. | Active |