Comment by wlonkly

> In other words, applications where some other system (human or otherwise) will catch the mistakes.

The problem with that is that when you move a human from a "doing" role to a "monitoring" role, their performance degrades significantly. Lisanne Bainbridge wrote a paper on this in 1982 (!!) called "Ironies of Automation"[1], it's impressive how applicable it is to AI applications today.

Overall Bainbridge recommends collaboration over monitoring for abnormal conditions.

[1] https://ckrybus.com/static/papers/Bainbridge_1983_Automatica...