One of the key factors for the acceptance of expert systems in real-world domains is the ability to explain their reasoning (Buchanan & Shortliffe, 1984; Henrion & Druzdzel, 1990). This paper describes the basic properties that characterise explanation methods and reviews the methods developed to date for explanation in Bayesian networks.

Explanation of reasoning is one of the most important abilities an expert system should provide in order to be widely accepted. In fact, since MYCIN, many expert systems have tried to include some explanation capability. This paper reviews the methods developed to date for explanation in heuristic expert systems.

Abstract-Bayesian networks and influence diagrams are probabilistic graphical models widely used for building diagnosisand decision-support expert systems. Explanation of both the model and the reasoning is important for debugging these models, for alleviating users' reluctance to accept their advice, and for using them as tutoring systems. This paper describes some explanation options for Bayesian networks and influence diagrams that have been implemented in Elvira and how they have been used for building medical models and for teaching probabilistic reasoning to pre-and post-graduate students.

No abstract

Student dropout in Engineering Education is an important problem which has been studied from different perspectives, as well as using different techniques. This manuscript describes the methodology used in order to address this question in the context of learning analytics. Bayesian networks have been used as they provide adequate methods for the representation, interpretation and contextualization of data. The proposed approach is illustrated through a case study about Computer Science (CS) dropout at the University of Castilla-La Mancha (Spain), which is close to 40%. To that end, several Bayesian networks were obtained from a database which contained 383 records representing both academic and social data of the students enrolled in the CS degree during four courses. Then, these probabilistic models were interpreted and evaluated. The results obtained revealed that the best model that fits the data is provided by the K2 algorithm although the great heterogeneity of the data studied did not permit the adjustment of the dropout profile of the student too accurately. Nonetheless, the methodology described here can be taken as a reference for future works.