This grant will support work on partial constraint satisfaction, both in theory and in the development of a constraint-based reasoning shell for building expert problem-solving systems. Partial constraint satisfaction requires relaxing mutually exclusive conditions until an approximate solution to the original problem can be found. The computer will accept constraint networks as input and will produce inferences by backtracking and relaxation (rather than by forward and backward rule chaining). The shell's interface will be a symbolic spreadsheet where the effects of changing constraints can be observed. Specific applications in wildlife management may be studied.