Collaborative Multi-Expert-Systems: towards more flexibly integrating and processing case-specific and (more) general knowledge

Klaus-Dieter Althoff
In: Andreas Henrich; Hans-Christian Sperker (Hrsg.). LWA 2013 - Lernen, Wissen & Adaptivität - Workshop Proceedings. GI-Workshop-Tage "Lernen, Wissen, Adaption" (LWA-2013), October 7-9, Bamberg, Germany, Uniuversity of Bamberg, Bamberg, 10/2013.


Case-based reasoning (CBR) and expert systems have a long tradition in artificial intelligence: CBR since the late 1970s and expert systems since the late 1960s. While expert systems are based on expertise and expert reasoning capabilities for a specific area of responsibility, CBR is an approach for problem solving and learning of humans and computers. Starting from different research activities, CBR and expert systems have become overlapping research fields. In this talk the relationships between CBR and expert systems are analyzed from different perspectives like problem solving, learning, competence development, and knowledge types. As human case-based reasoners are quite successful in integrating problem-solving and learning, combining different problem solving strategies, utilizing different kinds of knowledge, and becoming experts for specific areas of responsibility, computer based expert systems do not have the reputation to be successful at these tasks. Based on this, the talk will discuss the learning ability of expert systems on different levels and the role CBR may play here. A research project is introduced that aims at, among others, improving the learning ability of expert systems by systematically considering multiple expert(s) (systems) as well as the wisdom of the crowd. The corresponding software architecture integrates concepts from software engineering (experience factory, software product lines) and artificial intelligence (multi-agent systems, CBR). In the scope of this research CBR is used in various ways: for representing and processing the experience part of expertise, for supporting continuous knowledge evolution and increasing knowledge formalization, as well as for providing an open framework for constructing learning expert systems. The current state of implementation is presented as along with open challenges and an outlook on future research.

Weitere Links