Over the past few years, the application of expert systems has moved f
rom the domain of toy systems and menial tasks to areas where complex
systems perform critical tasks. However, the acceptance of expert syst
ems for practical use has been slow. This paper argues that one of the
reasons for this is that insufficient attention has been directed at
ensuring the correctness and reliability of expert systems in the earl
y phases of development. This article then discusses a life cycle appr
oach to identify and reduce the different types of reliability problem
s inherent in developing expert systems.