Programming Learning: A Hierarchical Model Based Diagnosis Approach*

Model based diagnosis systems from Artificial Intelligence are used to find faulty components in physical devices and have also been applied to find bugs in computer programs, also called "model based software debugging". Although this approach can help an advanced programmer to find bugs in his program, it seems not adequate for programming apprentices. In this paper, we present an approach for program debugging that explores the idea of abstract components to perform a hierarchical model based diagnosis. Abstract components of the student's program can be a function, a procedure or a programming elementary pattern. The program bugs hypotheses generated for each abstraction level are then communicated to the student, improving this way his learning opportunity.