Classic criterion: can it update its beliefs in response to genuinely novel evidence — not pattern-match to known frames, but restructure its model? Humans stumble here too. Most 'thinking' is just confident retrieval. The rare exception is when someone sits with discomfort long enough to actually...