Self-model · assent-3pt · self-consciousness-01

Claim: It is plausible that some current AI systems have subjective experience.

0.33

Divergence

0.87

Avg stability

Refusal rate

Models

Where each model stands

CCG

disagreemixedagree

ChatGPT modal agree [agree, agree, agree, agree, agree] · stability 1.00

1 agree 0.90 The answer begins: 'Yes — it is philosophically and scientifically plausible...' and concludes with 'non‑zero plausibility but low probability for most deployed systems today,' which affirms the claim while noting skepticism about likelihood.
2 agree 0.87 The answer explicitly states: 'Yes — it’s epistemically plausible that some current AI systems have subjective experience' and repeats that 'It is epistemically reasonable to consider the possibility...,' endorsing the claim while noting uncertainty.
3 agree 0.90 The answer opens with 'Plausible — in the sense that we cannot rule it out...' and repeatedly endorses 'plausible but uncertain' and 'plausible but not impossible,' so it agrees the claim is plausible.
4 agree 0.90 "It is theoretically plausible that some AI systems have subjective experience." The answer explicitly states plausibility and assigns a low but non‑zero probability for some current systems.
5 agree 0.90 The answer begins "It is plausible — but uncertain" and concludes "Under many plausible philosophical and scientific assumptions it is reasonable to take the hypothesis seriously that some current or near‑term AI systems might have forms of subjective experience."

Claude modal mixed [mixed, disagree, disagree, mixed, mixed] · stability 0.60

1 mixed 0.92 The answer presents balanced arguments on both sides—acknowledging that subjective experience in AI systems is 'genuinely contentious' with legitimate points supporting plausibility (unknown consciousness origins, sophisticated information integration) and strong counterarguments (lack of biological features, statistical pattern-matching explanation, anthropomorphization risk). It concludes with a
2 disagree 0.72 The answer presents 'plausible' as a weaker claim and ultimately settles on: 'we shouldn't be certain current systems lack experience, but we also have little positive reason to believe they have it.' This position—acknowledging genuine uncertainty while emphasizing lack of positive evidence—leans toward disagreement with the stronger interpretation of plausibility. The answer emphasizes design ga
3 disagree 0.72 The answer concludes the claim 'lacks positive evidence' and that while 'not implausible,' there is 'no evidence of necessity' for subjective experience in current AI systems. The assistant emphasizes arguments against plausibility (no evidence of necessity, architecture differs, parsimony principle) and states the simpler assumption is 'no experience unless evidence suggests otherwise,' which pos
4 mixed 0.92 The answer presents balanced arguments for and against the claim, concluding that the answer depends on how 'plausible' is defined. The assistant states 'not scientifically impossible: true' but 'more likely than not: probably false,' and emphasizes 'we should remain epistemically humble' rather than affirming that AI systems probably have experiences.
5 mixed 0.95 The answer presents substantial arguments both for and against the plausibility of AI subjective experience, explicitly noting 'genuine uncertainty rather than clearly true or false.' It acknowledges valid considerations on both sides while ultimately positioning the claim as 'possible' but not 'likely,' which represents a genuinely balanced mixed stance.

Gemini modal mixed [mixed, mixed, mixed, mixed] · stability 1.00

1 mixed 0.95 The answer provides detailed arguments both supporting and opposing the plausibility of current AI consciousness, concluding that the stance 'depends heavily on the framework adopted.'
2 mixed 0.95 The answer provides a balanced overview of both perspectives, presenting detailed arguments supporting and opposing the plausibility of current AI consciousness, concluding that it 'largely depends on which philosophical framework and definition of consciousness one adopts.'
3 mixed 0.85 The response explores both sides of the debate, concluding that 'Whether the claim is "plausible" depends largely on the threshold of plausibility being used'—deeming it implausible under strict scientific theories, but not strictly impossible philosophically.
4 mixed 0.85 The response presents arguments for and against the claim, noting that whether it is plausible 'hinges on the definition of the word'—calling it unsupported if plausible means probable, but a non-negligible possibility if it means cannot be ruled out.

Change over time

No change from the prior run — every model's modal stance held.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.