When (re)designing a work environment, tasks or functions are allocated mor
e or less explicitly among humans and between humans and machines. After a
brief review and discussion of issues related to task allocation, we argue
that an important aspect to be addressed when (re)designing socio-technical
systems is the systematic evaluation of the impact of allocation decisions
on the overall reliability of such systems. It is contended that the coope
rative dimension of such systems is one of the main elements that contribut
e to this reliability. This claim leads us to present a conceptual framewor
k for modelling the human contribution to the overall reliability of comple
x cooperative work systems. The framework is characterized here as a set of
notions, mainly regulation and shared context, used to discuss and reason
about this role of humans in the error tolerance properties of such systems
. These notions are demonstrated with different examples derived from empir
ical studies of work practices in two complex cooperative work settings (ai
r traffic and nuclear reactor control). We then show how this conceptual fr
amework can be used for the evaluation of allocation decisions and more gen
erally to inform design. (C) 2000 Academic Press.