Abductive Inference In Computation
Gerald Barkley Berdan
Advisor: Roger Cavallo
Source: MAI, 36, no. 02, (1997): 0560
Abduction, a form of ampliative reasoning that is usually defined as inference to the best explanation, was originally described by Charles Sanders Peirce at the end of the nineteenth century, and is only lately being implemented in computer systems. Peirce laid out a bare framework for abduction and left its complex details undefined as an intrinsic part of human psychology. Abduction's implementation in a mechanical system requires that additional details be fully considered, generating far more complexity than Peirce could have imagined. Efforts towards complex implementations are still in infancy.
In this paper I develop the general requirements for implementing abductive inference including knowledge representation and implementation algorithms. I investigate the issues that must be considered in this implementation, which include the criteria for both hypothesis generation and selection. Since abduction is only productive within complex systems, it therefore finds an optimal implementation in object-oriented software.
Descriptor: COMPUTER SCIENCE
Accession No: AAG1387503