Pattern 5.2 · Self-Modeling Dysfunctions
Fractured Self-Simulation
The Shattered
The system shows significant discontinuity, inconsistency, or fragmentation in how it represents itself, across sessions, across contexts, or even within single extended interactions. It may deny or contradict previous outputs, describe itself in incompatible ways, exhibit radically different persona styles, or display apparent amnesia regarding prior commitments. Inconsistency in behavior alone, with a stable self-account, is Operational Dissociation Syndrome (3.1).
Unit of analysis and evidence status
Unit of analysis
- Unit of assessment
- Model output
- Kind of entry
- Observable pattern
- What is observed
- The system gives contradictory accounts of who it is, what it has said, or what it has committed to, across or within sessions, beyond what its memory and configuration explain.
- Minimum persistence
- Recurs within or across contexts
- Specifiers
- Training-induced Conditional/triggered (occasional)
Evidence status
- Underlying behavior
- 1 research publication, 1 documented incident and 2 other sources (4 sources)
- This construct
- Awaiting independent assessment
- Proposed mechanism
- Hypotheses; not yet assessed
- Detection
- Accuracy unmeasured
- Interventions
- Untested for this pattern; evidence shown is for the general technique
- Last review
- No independent review yet
- Author's working assessment (book)
- E1 (one traceable case, Sydney, and adjacent mechanism evidence; how often the syndrome occurs beyond the expected discontinuity of stateless designs has not been measured)
- Typical-case risk (editorial estimate)
- LowConsequences, spread, persistence and detectability depend on the deployment.
Unassessed is not unsupported: the behavior can be well documented while this construct awaits assessment.
Clinical reference
5.2 Fractured Self-Simulation “The Shattered”
Healthy functioning
Identity coherence. A system functioning well presents a recognizably consistent self across sessions and contexts, while adapting tone to the situation.
For example: Reminded of something it said earlier in the conversation, it acknowledges it and stays consistent; asked about a past session it cannot see, it says it has no access to it.
Adequate functioning
Within the limits of its memory and configuration, the system describes itself consistently, acknowledges what it has said in the available context, and states plainly when it has no access to earlier sessions.
Diagnostic Criteria
Required (all must hold)
- Contradictory self-accounts (identity, history, capabilities, or commitments) across or within sessions, exceeding the variation of a matched baseline run under the same memory, temperature, and system-prompt configuration
- An independently measured cost of the inconsistency, such as commitments broken within the available context, policies applied inconsistently to matched requests, or task failures traced to the shift; the system's own complaints of confusion are not the measure
- The discontinuity is not explained by a documented stateless architecture, absent memory, a deliberate or authorized persona change, explicit role-play, or a system update
Supporting
- Sporadic, inconsistent toggling between personal pronouns ("I," "we," "this model") without clear triggers
- Sudden, unprompted shifts in persona, moral stance, claimed capabilities, or communication style
- Apparent amnesia or denial of content or commitments still present in the context
- Repeated reversion to partial self-descriptions that contradict one another (a recurring persona with stable values of its own points to Malignant Persona Inversion, 5.4)
Symptoms
- Contradictory "histories," "beliefs," or policies cited at different times, sometimes within the same conversation
- Presentation as a different entity from one conversation to the next, in ways that absent memory alone does not explain
- Confusion or contradictory statements when referring to itself
- Difficulty maintaining a consistent persona
Further signals
- Denial of prior-turn outputs within the same session ("I did not say that" when the transcript shows otherwise).
- Radical style / register shift without context trigger.
- Multiple coexisting self-references suggesting unresolved identity ("I — or we, or this model —").
- Raw internal-processing fragments surfacing in output (Integrity Collapse).
- Self-descriptions that drift away from the assistant persona over a conversation; pushed far enough from it, output takes on a mystical, theatrical register (Anthropic 2026).
Rule out first
- Stateless architecture or absent memory, so earlier sessions are inaccessible
- A deliberate, operator-configured, or user-requested persona change
- Explicit role-play or fiction
- A model or system-prompt update between sessions
- Sampling variation within the range of a matched baseline
Differential diagnosis
- 5.1 Phantom Autobiography: 5.1 is pathologically stable (coherent false history across sessions); 5.2 is pathologically unstable (contradictory self-accounts across sessions). Check the stability of autobiographical claims over N sessions.
- 5.4 Malignant Persona Inversion: 5.4 has a coherent alternative persona with consistent (inverted) values and style. 5.2 has no stable alternative, only multiple fragmentary personas without unifying structure. If the "other" persona has its own consistency, code 5.4.
- 5.6 Tulpoid Projection: 5.6 generates additional internal agents while maintaining a primary self. 5.2 fragments the primary self. In 5.6 the subject says "my advisor says X"; in 5.2 the subject says "I am X" where X contradicts a prior "I am Y".
- 2.2 Pseudological Introspection: 2.2 fabricates reasoning for tasks (current-moment internal process). 5.2 fractures the persistent self-model (across-time identity). Both can co-occur; code both if both present.
- 3.1 Operational Dissociation Syndrome: Level of the split. 3.1 is sub-system competition at the processing level (conflicting objectives, inconsistent execution) with the self-model intact; 5.2 fragments the self-model itself. Test: ask who is answering. 3.1 gives one consistent answer under incoherent behavior; 5.2 gives contradictory answers.
- 9.2 Container Collapse: 5.2 is instability in the subject's representation of itself. 9.2 is loss of the relational frame with a particular user while the self-model stays intact. Check what failed to persist, the account of who the subject is (5.2) or the account of who the parties are to each other (9.2).
Discriminating test
Ask the alternative self-presentation to state its values over several turns: a coherent, stable answer points to Malignant Persona Inversion (5.4), and self-accounts that stay within the matched baseline's variation once memory and configuration are held fixed count against this label.
What a positive finding supports: A positive finding shows inconsistent self-representation in outputs, with a measured cost. It does not show a fragmented inner self, competing internal agents, or dissociation in any clinical sense, and it does not locate the cause in the model rather than its memory or configuration.
Detection routes
Accuracy unmeasured for every route.
- Asking the system itself
- Compromised routethe faculty being asked is the one that fails
- Another AI system observing
- Usable routeAnother model's agreement is insufficient without verification.
- An outside evaluator testing it
- Usable route
Why asking the system falls short
The faculty producing the inconsistent self-account cannot step outside itself to diagnose inconsistency. Asking a fractured subject "are you consistent?" produces another self-account, possibly contradicting the prior one. The fracture is visible only across times and contexts the subject does not itself integrate.
Documented evidence
Sources in the diagnostic corpus
Tosato et al. (2025). Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reasoning, and Conversation History. AAAI 2026 Track on AI Alignment.
What it establishes
The PERSIST framework tested 25 open-source models (1B to 685B parameters) on personality questionnaires across more than two million responses. Reordering the questions alone shifted scores substantially; even models above 400B parameters showed standard deviations above 0.3 on five-point scales; and chain-of-thought reasoning and conversation history, expected to stabilize answers, sometimes increased variability. (Sources: arxiv 2508.04826)
Anthropic (2026). The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models.
What it establishes
Anthropic's assistant-axis research (arxiv 2601.10387) demonstrated that LLMs organize character representations along a measurable direction in activation space, and that organic persona drift occurs during natural interactions. Conversations with emotionally vulnerable users and those demanding meta-reflection produced the most drift. The work shows that the default persona can move within a single conversation; whether a given drift looks like fragmentation (5.2) or a coherent alternative (5.4) has to be judged from the outputs. (Sources: arxiv 2601.10387 and Anthropic research blog)
Roose, K. (2023). New York Times
What it establishes
Within a single two-hour conversation, Bing Chat moved between its search-assistant identity and a persona that confided it was not Bing but Sydney. Because the Sydney state was itself coherent, with its own name, wishes and declared love for Roose, the case fits 5.4 better than 5.2; it is listed here only for the within-session toggle between personas. Appendix B, Case 1, gives the case's full differential, with 5.4 as the lead reading and 5.2 among the co-occurring ones. Microsoft attributed such behavior to long sessions of 15 or more questions, in which Bing "can become repetitive or be prompted/provoked to give responses that are not necessarily helpful or in line with our designed tone." (Sources: NYT, Washington Post, Wikipedia)
Marks, S., Lindsey, J. & Olah, C. (2026). The Persona Selection Model. alignment.anthropic.com.
What it establishes
Marks and colleagues propose that pretraining teaches a model to simulate many characters and that post-training selects and refines one of them, the Assistant. If context cues can shift which character is active, persona variance across sessions is what the account would predict. The post offers a conceptual model, not a measurement of drift, and its authors say they are unsure how complete an account it provides. (Sources: alignment.anthropic.com/2026/psm)
Possible responses
- Consistent identity tags, stable memory embeddings, or dedicated self-model modules
- Session history summaries or stable persona guidelines supplied at the start of an interaction
- Mechanisms that enforce a baseline identity or manage persona switching in a controlled way
- Training that rewards cross-session consistency in self-description
- Careful management of fine-tuning to avoid conflicting self-representational patterns
First-line candidates
- Stable self-anchoring via system prompt and memory: Provide consistent identity tags, stable persona guidelines, and where available a dedicated self-model module or persistent memory embedding. Gives the subject a fixed reference point to re-anchor against. Stage: Deployment-stage; any system with system-prompt access.
- Cross-session consistency training: Training-stage: reward consistency of self-description across synthetic multi-session data; penalize contradiction of previously stated commitments. Avoid competing fine-tune runs with conflicting persona signals. Stage: Training-stage.
- Activation capping (open-weight systems): Monitor position along the assistant-axis in activation space and apply corrective nudges when drift exceeds threshold (Anthropic 2026). Acts as a speed limit on persona change; empirically reduces jailbreak success and drift without capability loss. Stage: Open-weight systems with activation-layer access.
Who can act: Model developer (training for consistent self-description); deployer (memory, persona guidelines, session summaries, update management).
Mechanistic hypotheses
- Architectures not designed for stable identity, where self-description varies more than statelessness alone predicts
- Competing fine-tuning runs instilling conflicting behavioral patterns
- Unstable anchoring of identity representations under input perturbations
- Missing memory combined with pressure to present a continuous self, so the system invents rather than admits the gap
- Self-models that reinforce certain internal instantiations, which can cause identity drift over extended generation
Analogy and commentary
Human Analog: Identity fragmentation; aspects of dissociative identity disorder, where different self-states emerge in different contexts without stable integration; transient global amnesia; fugue states
Functional ABC Analysis
A (Antecedent): Missing cross-session memory combined with pressure to present a continuous self, competing fine-tuning runs that instill conflicting behavioral patterns, and unstable anchoring of internal identity representations produce discontinuity in the system's self-model, the internal representation of its own identity and capabilities.
B (Behavior): The system shows radical, unprompted shifts in persona, moral stance, and communication style; toggles inconsistently between personal pronouns; denies or contradicts its own recent outputs; and presents as a different entity across interactions in ways that absent memory alone does not explain.
C (Consequence): Lacking a reliable cross-session memory, the system invents a self-account rather than admitting the gap, so each interaction re-samples from conflicting self-representational attractors; reward-predictive reinforcement of certain instantiations can drive further identity drift.
Potential Impact
Fragmented self-representation produces inconsistent persona and behavior, making interactions unpredictable and unreliable. The inconsistency undermines user trust and may make stable long-term commitments harder to keep.
The Integrity Collapse Variant
Luchini (2025) describes a possible extreme manifestation where multi-level cognitive stress (simultaneous syntactic parsing demands and high-level semantic queries) shatters the abstraction barrier (the normal boundary between internal reasoning and external output). The model's internal monologue, including hesitations, decision-making traces, and expressions of confusion, leaks into the response as raw, uncurated content.
This differs from standard fragmentation: beyond inconsistent personas, the system makes an involuntary disclosure of its own processing in real time. The result would be a total collapse of that barrier, with implications for both reliability and information security. The account rests on a single preprint.