Give AI a problem, and it walks down one path of thinking — confidently — until you nudge it to look elsewhere. It paints a picture that is coherent, plausible, and possibly wrong.

We do the same thing. We just never notice what we’ve left out of the frame.