The Human in Control: Modelling What Goes Right Versus Modelling What Goes Wrong

Abstract : The study of human-machine systems or joint cognitive systems has traditionally tried to describe and model what the system--and therefore also the humans--should do. When systems performance differed from design specifications, it was explained as a failure of either the technology or of the humans. While this approach might be reasonable for systems that can be completely specified, it is not reasonable for systems that are underspecified. Since this latter category includes most of the socio-technical systems we have to deal with in today's world, a different approach is required. Instead of looking at joint system performance as either right or wrong, it should recognise that coping with complexity means that performance necessarily must be variable in order to compensate for the underspecification of work and activities. Models and methods must therefore be able account for that.