STAHLp: Belief Revision in Scientific Discovery

III this paper we describe the STAlILp system for inferring components of chemical substances i.e., constructing cornponential models. S’l‘AIil,p is a descendant of the S’I’AI-11, system (Xytkow & Simon, 1986); both use chemical reactions and any known models in order to construct new rrlodels. I lowever, S’I’A 11 Lp employs a more unified and effective strategy for preventing, detecting, and recovrbri rig from errorleous inferences. This strategy is based partly upon the assumption-based method (de Kleer, 1984) of rccordillg the source beliefs, or premises, which lead to cacti illfclrred belief (i.e., reaction or model). STAlI 1,‘s multiple methods for detecting and recovering frorn erroneous inferences have been reduced to one method in STAH Lp, which can hypothesize faulty premises, revise them, and proc& to construct new models. The hypotheses made during belief revision can be viewed as interpretations from comptating theories; how they are chosen thus determines how ttieories evolve after repeated revisions. We analyze this issue with an example involving the shift from ptllogiston to oxygeii theory.