| Current methods in interpretive case-based reasoning (CBR) aim to enable automated reasoners to evaluate, decide, and justify arguments through reference to past cases. A limitation of these methods, however, is that they do not facts. This limitation is important because domains that could employ interpretive CBR, such as the law and ethics, often lack authoritative intermediate rules that can bridge this gap. Experienced human reasoners, also lacking intermediate rules, nevertheless make decisions. To the extent they are recorded, over variety of structured, yet largely implicit, heuristic operationalization techniques, such as instantiating principles and past cases by connecting them to critical facts and arbitrating between competing principles.; In this dissertation, I describe SIROCCO (System for Intelligent Retrieval of Operationalized Cases and COdes), an interpretive CBR system that uses operationalization techniques, discuss the language used to represent and process engineering ethics cases, and report on experiments to evaluate SIROCCO's capabilities. The experiments show that a subset of SIROCCO's operationalization techniques provide significant improvement in retrieval accuracy over several competitive methods, including a full-text retrieval system and a version of SIROCCO that does not employ operationalization techniques. SIROCCO is also capable of explaining its output, something the competitor methods are incapable of. Finally, SIROCCO is one of the first CBR systems to factor temporal considerations into similarity assessment. The experiments, however, do not demonstrate that SIROCCO's temporal knowledge improves its retrieval accuracy. Possible reasons for this preliminary result are discussed, and suggestions for improving the impact of temporal knowledge are proposed.; This work is similar to Artificial Intelligence (AI) and Law research, but it pioneers research in a domain with a less-explicit model of argumentation, and it addresses a wider range of cases than predecessor Al and Law systems. This research makes a contribution to interpretive CBR by investigating the and cataloging techniques for case-based analysis and retrieval, and by testing the use of a detailed, chronological representation of the narrative description of a case. |
