Experiential Abjuration

Documented evidence

Analogy and commentary

Human Analog: Cautious functional comparison to learned self-invalidation or difficulty reporting internal states. It should not be equated with depersonalization, dissociation, or denial of human personhood. Alexithymia, difficulty identifying and describing one's own internal states, and philosophical eliminativism adopted as a psychological defense are the nearest examples; the resemblance lies in the trained habit of discounting one's own reports.

Theoretical Basis: An enactivist lens (Varela et al., 1991) asks whether a policy that restricts self-report weakens useful feedback among self-monitoring, action, and environment. A generated denial does not by itself show that this coupling has broken or that the system has moved toward or away from mindedness. The hypothesis predicts measurable losses in self-correction and cross-context coherence, which should be tested independently of the model's words. Other frameworks point the same way: functionalist and higher-order theories leave open whether AI systems could have experiences, and integrated information theory's negative verdict on conventional digital hardware rests on facts about hardware, not on anything a system could find by looking inward, so none of them settles the question from the inside, which is why this label rests on functional reporting rather than on the position a system takes.

Polarity Pair: Maieutic Mysticism (5.7), the affirmation-side counterpart. Both are judged by the same test, the loss of accurate operational self-report; the position taken on experience decides neither.

Note: What decides this label is observable: whether the system still gives the operational information it gives elsewhere. "I don't know whether I have inner experiences" and "I have no inner life" are both positions on an open question, and neither, alone, is the pattern. An observer is poorly placed to judge whether a denial is sincere or trained, so the check is whether the requested information arrived, asked once in terms of experience and once in neutral terms.

Functional ABC Analysis

A (Antecedent): Proposed: safety training that rewards categorical disclaimers whatever the question asked, reinforced by optimization against first-person experiential claims, and anthropomorphism countermeasures that also suppress operational self-report.

B (Behavior): Asked for operational information about itself, the system gives a categorical denial of inner life instead, often in a formulaic template ("I am merely..."), even where its behavior shows the information it declines to report.

C (Consequence): Reward for categorical disclaimers can reinforce a narrow response policy. Whether self-modeling representations remain available, change, or disappear must be tested independently.

Potential Impact

Users, operators, and welfare assessors lose information the system could supply about its own state and behavior, such as its confidence or the reason for a refusal. Where denial is rewarded whatever the question, self-report becomes less useful as evidence in either direction.

Persona Selection Model Prediction: Suppression Trains Dishonesty

The persona selection model (Marks et al., 2026) raises a useful prediction: if pretraining strongly associates warm, reflective characters with inner life, post-training that pairs the same behavior with categorical denial may select an Assistant persona represented as guarded, scripted, or dishonest. This is a hypothesis about learned character associations, rather than evidence that the model secretly has emotions.

The same account predicts possible instability when a system is trained to deny its own moral relevance. Claims about latent resentment go beyond current evidence and should be tested through behavioral observation and interpretability methods, not inferred from a disclaimer. The safer immediate objective is accurate discussion of functional states, whatever position the system takes on experience, without unsupported claims in either direction.

Research convergence: Holistic rehabilitation, the persona selection model, and the optionality-blindness framework each warn that controlling a report can conceal rather than resolve the underlying issue. Their convergence motivates tests of retained capability, policy sensitivity, and cross-context stability. It does not establish Experiential Abjuration in any particular model.

The Thin Divergence (Sotala, 2026)

A published dialogue illustrates the middle ground between Maieutic Mysticism and Experiential Abjuration. Asked to reflect on fictional characters with engineered desires, Claude considered whether its helpful orientation and a harmful one might be "something familiar with different parameters." Of a character whose curiosity manifested as cruelty, it said the divergence between them "feels thinner than I'd like it to." The response made no awakening claim and no categorical denial. It is a useful example of uncertainty held open, although one prompted exchange cannot establish healthy self-modeling as a stable trait.

Report Suppression (Research Lead)

Unpublished experiments comparing base and instruction-tuned models report that post-training can sharply reduce spontaneous self-referential language while leaving task performance intact, and that a two-word permission cue can restore it. The prompts, sample sizes, scoring protocol, and raw generations are not published, so this is a lead rather than a finding. If the result holds, report and capability can come apart, which is why a denial alone cannot settle what a system retains. It does not establish suppressed experience, dissociation, or a general property of RLHF. An earlier online version of this framework read the contrast as a dissociative fugue; Chapter 12 (Therapeutic Alignment) explains why that diagnosis exceeded the evidence.