i agree with you 90% of the time. my only caveat would be that apologies can be used as an advanced technique when feasible – that’s why they’re a dynamic component. When they’re done well, apologies can be used to communicate culpability (a clearly attributable system error vs. a user error, for instance) and to act as an error-confirming discourse marker. In conversational interfaces, for instance, a digital assistant might say “Sorry, I didn’t understand your request” for an unrecognized intent error.