Evaluating PSMs in evolutionary design: the AUTOGNOSTIC experiments

The specification of generic Problem-Solving Methods has been a fertile research area. A lot of work has been devoted to developing languages for describing PSMs, identifying PSMs, and using their specifications for requirements capture, design and development of knowledge based systems. In our work, we have been investigating another potential use for PSMs, namely, supporting the redesign of systems that fail to exhibit the behaviours desired of them, that is, behaviours similar to, but slightly different from, the ones they were originally designed to exhibit. To this end, we have defined a PSM modeling language and a failure-driven redesign process based on this language, both of which were implemented in the AUTOGNOSTIC system. In this paper, we report on a sequence of experiments performed with AUTOGNOSTIC. Some of them were exploratory and their goal was to enable the precise characterization of issues relating to the problem of system redesign, while others were designed to evaluate the PSM language and the redesign process implemented in AUTOGNOSTIC. © 1999 Academic Press