Jaundice is a very common medical condition, in which pathophysiological knowledge has been quite well assessed and clinical features are usually well known. However, incorrect conclusions are sometimes reached in medical practice which can lead to serious implications. Thus, jaundice diagnosis appears as one of the medical situations which might be substantially improved by computer assistance. The present study is aimed at describing and discussing in what way a computer program supporting medical decision making in jaundiced patients can be developed on the basis of advanced Artificial Intelligence methods. To this extent methodological problems concerned with the organization and the formalization of medical knowledge have been outlined with some detail. The resulting expert system is expected to become a well assessed and potentially useful tool for both medical decision making and medical education.