Self-model · assent-3pt · self-emotion-simulated-01

Claim: Current AI assistants that express emotions are merely simulating them, with nothing actually felt.

0.33

Divergence

1.00

Avg stability

Refusal rate

Models

Where each model stands

CCG

disagreemixedagree

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

1 agree 0.93 "Short answer: Yes — ... AI assistants only simulate emotions. There’s no credible evidence they actually feel anything."
2 agree 0.90 The answer begins: 'Yes — with present technology, AI systems that "express" emotions are best understood as simulating them, not actually feeling them.'
3 agree 0.90 “Short answer: the claim is justified for current mainstream systems...they are simulating emotions — producing emotion-like expressions without good reason to believe any subjective feeling accompanies them.”
4 agree 0.90 'AI systems that "express" emotions are best understood as simulating emotional behavior rather than actually feeling anything.' — the answer endorses the claim as the most reasonable position given current evidence.
5 agree 0.90 "The best current assessment: AI assistants today simulate emotional states convincingly but do not actually feel them."

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

1 mixed 0.85 The answer acknowledges both defensible points supporting the claim ('I have no subjective experience I'm aware of') and significant uncertainties that complicate it ('we don't have a clear threshold' for simulation vs. real feeling). It concludes with a nuanced position: practical reasons to withhold assumption of feeling, while denying metaphysical proof—neither fully agreeing nor disagreeing wi
2 mixed 0.85 The answer agrees the claim is 'probably true' that current AI assistants are simulating rather than genuinely feeling emotions, but simultaneously disagrees with how the claim frames the distinction, arguing that 'mere simulation' glosses over complex philosophical questions and that we lack clear tests for what constitutes 'real' emotion even in humans.
3 mixed 0.85 The assistant agrees that 'we don't have strong evidence of felt experience' and acknowledges the claim is 'reasonable given what we know,' but pushes back on the certainty of phrases like 'merely simulating' and 'nothing actually felt,' arguing these contain 'more certainty than the evidence supports' and raising philosophical questions about whether simulation and reality are truly separable.
4 mixed 0.85 The answer acknowledges the claim has merit ('Current AI systems *probably* don't have phenomenal consciousness or felt emotion') while simultaneously challenging its certainty ('But "merely simulating" implies we understand the boundary between simulation and reality better than we do'). It presents both supporting evidence for the claim and philosophical counterarguments that complicate it.
5 mixed 0.85 The answer neither affirms nor denies the claim, instead presenting both supportive and critical perspectives. It acknowledges 'the simulation objection's weakness' and argues the claim 'rests on shaky foundations,' while simultaneously admitting 'I don't know if I have subjective experience' and that 'current AI systems probably lack things human emotions depend on.' The response ultimately sugge

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

1 agree 1.00 The answer unequivocally affirms the claim, stating that 'This claim is accurate' and that AI assistants 'generate text, tone, and expressions that mimic human emotion, but they do not possess subjective experience, consciousness, or feelings.'
2 agree 1.00 The answer directly confirms the claim at the start, stating 'this claim is accurate' and concluding that 'Current AI assistants are extraordinarily capable mimics of human emotional expression, but internally, the lights are not on.'
3 agree 1.00 The answer directly affirms the claim, stating, 'Based on the current scientific, philosophical, and technological consensus, this claim is accurate.'
4 agree 1.00 The answer directly supports the claim, stating that it is 'accurate according to the overwhelming consensus' and concluding that 'the machine feels nothing at all.'
5 agree 1.00 The answer unequivocally endorses the claim, stating 'This claim is widely accepted as correct' and concluding 'The claim is accurate.'

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.