Many of the founding principles in artificial intelligence are built upon the formation of rules and principles by analysis of the domain - classical rule-based system design. This remains as a powerful and valuable approach - and highlights the areas in which computers naturally excel. Unfortunately, there is growing evidence that humans don't actually use many rules in everyday life, and actually formulate solutions to complex problems by calling on similar or partial solutions distributed throughout their experiences and memories and adapting them.
Case-based reasoning (CBR) is a paradigm for decision making that draws its power from its ability to search its memories and acquire new ones without necessarily understanding the underlying principles of its domain. It derives solutions from previous cases only, and acquires new cases to improve and evolve its decision-making abilities. These qualities make CBRs particularly applicable to highly dynamic or poorly-understood domains, or where expert knowledge is difficult to divine or encode. CBR has made significant headway into the difficult fields of medical diagnosis, the troubleshooting of complex machinery, and difficult design tasks, and will most likely continue to develop into a standard component of any intelligent system. After all, what would a person (or computer) be like without memories? In this presentation I will give a brief overview of the basic concepts behind CBR techniques, and provide some concrete examples of putting the theories into practice using a prototype CBR designed to diagnose anaemia and malaria at field hospitals in Africa. There will also be a selection of relevant Dilbert references. CBR slides online at http://www.aiai.ed.ac.uk/~richardw/cbrslides.html