Integrating research into the core curriculum. A case study

Summary form only given. This presentation describes an approach to providing undergraduates with a "real-world" research experience within the context of a core computer science course. As a required course, Introduction to Artificial Intelligence (AI) is typically taught in the junior/senior year, and normally covers fundamental topics in AI as well as Lisp programming. The author wanted to give the students an experience of what is involved in a real AI research project, such as the Visual Interaction Assistant for Radiology. This intelligent assistant is based on extensive cognitive data collection from radiologists reading chest X-rays, and there are several open analysis problems which still need to be resolved. It seemed that the "think aloud" reports of the radiologists could be used as a vehicle to help the students learn how to extract computationally meaningful information from a domain expert's thoughts and words. Many issues arising in this exercise are applicable to the development of expert systems in other domains as well, for example: handling specialized vocabulary, developing quantitative measures and qualitative concept labeling procedures, applying subjective ratings, aggregating data, and interpreting results of the analyses both individually and in the context of the whole experiment. The required tasks included a time analysis using the original video data, a concept analysis where the words in the protocols were categorized and labeled, and a performance analysis where the subjects' diagnoses were compared with those of the project expert.