Human plausible reasoning for intelligent help

This paper is about providing intelligent help to users interacting with an operating system. Its main focus is an investigation of Human Plausible Re asoning Theory (Collins & Michalski, 1989) to infer the commands the user s hould have typed, given what they did type. The theory has been adapted and incorporated into a prototype Intelligent Help System (IHS) for UNIX users , called RESCUER, and has been used for the generation and evaluation of hy potheses about users' beliefs underlying the observed users' actions on the UNIX file store. The hypotheses generated by RESCUER were compared to thos e made by human experts on the sample scripts from UNIX user sessions. The potential for Human Plausible Reasoning as a mechanism to reason about slip s and misconceptions is discussed.