A Review of Explanation and Explanation in Case- Based Reasoning 1 Dónal Doyle, Alexey Tsymbal, Pádraig Cunningham Department of Computer Science Trinity College Dublin Donal.Doyle@cs.tcd.ie 1. Introduction With the arrival of more powerful computers and improved algorithms in machine learning, the availability of computer-based Knowledge Based Systems (KBSs) has increased rapidly. There is a wide range of KBSs available in many different types of domains, including medicine, finance, industry and technical diagnosis (Armengol et al, 2000; Ong et al, 1997; Rowe & Wright, 1993; Mark et al, 1996; Ye, 1995). Sales of KBS development tools are growing at a rate of about 16% per annum since 1988 (Durkin, 1996). For example, the French Bank “Evalog” has decreased its costs of processing loans by tenfold, helped minimise risks and increased processing capacity by using a KBS called “EvEnt” (Mao and Benbasat, 2000). However, case studies show that, in contrast to the availability of these systems, many KBSs that are installed in organisations are not being used (Majchrzak and Gasser, 1991). Some possible reasons for the lack of usage of available KBSs is that the users do not understand the models used in the particular system (Majchrzak and Gasser, 1991), or they might not be convinced by the prediction given by the system (Brézillon and Pomerol, 1996). Other possible reasons could include a fear by system users that the system will eventually replace the user (Berry and Hart, 1990), or a fear that the use of KBSs may lead to a dehumanisation of some work practices (Shortliffe, 1992). Consider one possible scenario occurring from the use of KBSs in the medical domain of the future. “The semiconscious patient lies in a futuristic intensive care unit, tubes protruding, wires emerging from under the sheets and connecting to a host of monitor carts or wall-mounted devices, and intravenous fluids with computer-controlled infusion pumps circling the bed. The beeps of the monitors are not interrupted by footfalls of nursing staff, for health workers seldom have to enter the room. Instead, intelligent devices measure every pertinent physiological parameter, deciding how to adjust infusion rates, when to alter the respirator settings, and whether to sound alarms for the intervention of nurses or physicians." (Shortliffe, 1992) To some people this scenario for the clinical use of computers may appear unrealistic, due to current limitations of automation and ethical or legal constraints 1 This material is based upon works supported by the Science Foundation Ireland under Grant No. S.F.I.-02IN.1I111.