Lots of difficult questions arise from his, and others work. Is it logical, for example, on a flight deck, to allocate the task of monitoring the aircraft systems' performance to a human, and then expect them to 'jump in' and resolve any issue that may occur, rather than have an automated system monitoring the human performance, and alerting when something is missed?
Being on the ground-based side of aviation, I am constantly reminded of this dichotomy when hearing of future plans for controllers to simply monitor flights, all of which are all following PRNAV routes knitted through the sky. The idea being, we would simply be a 'fall back', ready to jump in and resolve a situation which may have developed. Personally, I think my chances of being able to help if such a situation arose, having been diligently 'monitoring' the situation without intervention for perhaps a year or so, are close to zero.
Virtually all evidential testing seems to lean towards humans being rubbish at monitoring systems for perhaps initially minor unexpected changes, and automated systems being not much better at producing instant dynamic plans in reaction to unfamiliar scenarios, and yet that seems to be the path we have chosen to proceed along.
Fortunately, I will be retired before the controller task is reduced to such a state, and I'm sure systems will arrive that will do a grand job, for those 99.99% of the times when things are going swimmingly. I really don't fancy being the one to have to try to help when something does go wrong though.