No, I mean over-update in a Bayesian capacity. As in, they are updating their prior probability using evidence to arrive at the posterior probability.
Think text much more likely from robot than first thought
Grug say this change too big from just one em dash