In Bayesian terms, stupidity is sufficiently abundant that even fairly strong evidence of harmful intent can’t overcome the base rate.
I interpreted this as both this effect that you described and that the harm of a false assumption of malice is greater than that of a false assumption of stupidity, so expected utility favours acting as if people were stupid rather than evil.
I interpreted this as both this effect that you described and that the harm of a false assumption of malice is greater than that of a false assumption of stupidity, so expected utility favours acting as if people were stupid rather than evil.