From: Nancy Leveson (leveson(at)sunnyday.mit.edu)
Date: Tue 19 Nov 2002 - 11:10:31 GMT
The basic principle here, in my experience, has to do with the effectiveness of monitors in general. Human monitors are very ineffective if the thing (either human or automation) being monitored rarely makes mistakes. Alternatively, humans who know they are being monitored (either by another human or an automated system) tend to rely on the monitor and turn their attention to other things. That's because we move toward efficiency in our jobs. So what starts out as redundancy turns into a system without much redundancy. As has been noted by others in this interchange, a more careful task analysis where each person has assigned tasks that together add up to a safer system is required. Simply treating people like hardware and naively adding another monitoring "component" won't work. Nancy