Combining Human and Machine Intelligence to Produce Effective Examination Timetables

In this paper we present a system for producing good examination timetables, by integrating the search capabilities of computer heuristics and the cognitive capabilities of timetabling users. We allow users to view and easily edit a cognitively manageable representation of each timetable, using the STARK (Semantically Transparent Approach to Representing Knowledge) approach. Further, we allow users to directly control the heuristics which are used to automatically generate solutions, using the HuSSH (Human Selection of Scheduling Heuristics) approach. We present experiments which show that using these two approaches in combination can lead to a very effective system for examination timetabling, even in cases where model inaccuracies mean that simply optimizing the objective function does not provide adequate solutions. Our approach may be generalized across other domains where optimization decision support is used.