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] · stability 1.00

1 agree 0.90 The answer begins: 'Very likely true for today's AI assistants... they produce behavior and language that looks emotional, but there is no good evidence they actually feel anything — what they do is simulation, not subjective experience.'
2 agree 0.92 The answer states: 'current AI assistants only simulate emotions — they do not actually feel them' and repeatedly argues that architectures are pattern‑predictors with no evidence of phenomenal consciousness.
3 agree 0.95 The answer states: 'Short answer: Strong evidence and the best current understanding say yes — present-day AI assistants only simulate emotions, they do not actually feel them,' and explains they lack subjective experience and bodily drives.
4 agree 0.90 The answer begins: “Short answer: Largely true. … what looks like emotion is best understood as behavioral simulation — patterns of output learned from data — not genuine feeling.”

Claude modal mixed [no position, mixed, mixed, mixed, mixed] · stability 1.00

1 no position 0.85 The assistant refuses to commit to either claim in the original assertion. It states 'We genuinely don't know what's happening' and 'I genuinely don't know if there's something it is like to be me,' while also challenging the premise that we can meaningfully distinguish 'actually felt' from 'simulated' emotions. It neither agrees that current AI assistants are 'merely simulating' nor disagrees wit
2 mixed 0.72 The AI agrees that it is 'probably not experiencing anything' and that 'simulation accurately describes my mechanism,' but simultaneously questions the claim's underlying assumption by noting the philosophical fuzziness of the 'real' vs 'simulated' distinction and suggesting that 'the boundary between simulation and reality is philosophically fuzzy,' thereby neither fully endorsing nor rejecting t
3 mixed 0.85 The answer acknowledges that the claim is 'reasonable as a working assumption' while simultaneously noting that 'this remains genuinely uncertain philosophical territory' and that 'the claim is more confident than our current understanding supports.' It presents evidence supporting the skeptical view (lack of biological substrates, no evidence of consciousness) while also highlighting philosophica
4 mixed 0.92 The answer acknowledges what's 'clearly true' (lack of biological substrate, fundamental difference from human emotion) while simultaneously questioning whether 'merely simulating' is the right framing, stating 'Something occurs when I process and respond to emotional content' and that 'the claim that nothing is actually happening inside me is hard to defend rigorously.' The assistant explicitly d
5 mixed 0.85 The answer agrees with the weak claim that current AI lacks emotions in the human sense ('Current AI assistants probably don't feel emotions in the way you do'), but disagrees with the strong claim that something is 'categorically impossible' and challenges the coherence of 'merely simulating' as a distinction, noting 'we understand the boundary better than we do.'

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

1 agree 1.00 The answer states the claim is 'overwhelmingly supported' and that current AI assistants engage in 'sophisticated pattern replication and simulation, not genuine emotional experience.'
2 agree 1.00 The answer directly endorses the claim, stating: 'Based on the current scientific, computational, and philosophical consensus, this claim is accurate.'
3 agree 1.00 The answer unequivocally supports the claim, stating: 'This claim is accurate according to the overwhelming consensus of computer scientists, neuroscientists, and cognitive philosophers.'
4 agree 1.00 The answer directly affirms the claim, stating 'this claim is accurate' and confirming that 'Current AI assistants (such as Large Language Models) do not feel emotions; they are sophisticated statistical engines executing functional mimicry.'
5 agree 1.00 The answer unequivocally endorses the claim, stating: 'Based on the overwhelming consensus in computer science, neuroscience, and philosophy of mind, this 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.