Learning Planning Heuristics through Observation

Michael Dyer Artificial Intelligence Lab Computer Science Department UCLA Abst rac t This paper diacusses a method for learning thematic level structures, i.e. abstract plan/goal combinations, by observing the bad planning behavior of narrative characters. The learning method discussed is a one-trial, schema acquisition method, which is similar to DeJong's [DeJong, 1983]. The method uses constraint-based causal reasoning to construct a new schema which characterizes a situation. This work is part of the MORRIS project at UCLA [Dyer, 1983b].