A Framework for Abductive Computational Modeling

Abstract

This Research to Practice full paper presents a framework for abductive computational modeling. Abduction is a type of reasoning which starts with observed facts then seeks to find hypotheses to explain the facts. This type of reasoning is fundamental for engineers to solve many problems, including design, planning, and fault diagnosis. Computational modeling is the process of programming models of machines, circuits, buildings, and phenomena in general. Abductive Computational Modeling refers to development of programs to make abductive reasoning about devices, techniques, and phenomena that are commonly studied, for example, in the curricula of engineering courses. As an educational activity, modeling belongs to the class known as constructivism, which students learn through activities of developing models and critically analyzing the developed model. Although computational modeling has been extensively investigated–-since Papert’s seminal studies in the 1960s about the use of the LOGO language, until the present day, demarcated by modeling with Arduino, Raspberry Pi, and languages like C, Scratch, and Python–-there is little study on abductive computational modeling. This work proposes and experimentally investigates the use of a framework for abductive computational modeling named AbCM. The results suggest the feasibility of the framework and indicate important challenges to be overcome.