Novices’ conceptions of machine learning
Andreas Mühling, Gregor Große-Bölting
Abstract
With machine learning becoming more and more prevalent in society, it also becomes a topic that is relevant for general K12 education. At the same time, modern machine learning is relying heavily on mathematics, requiring novel approaches to teaching it in settings where a full coverage is not possible. In this context the study presented here investigates the conceptions that students have about the workings of a particular type of machine learning system before and after a short workshop. Students (N = 57) gave open ended answers that were analyzed qualitatively, following the approach of phenomenography. The resulting model was then validated with semi-structured interviews with teachers (N = 5). The results indicate that students’ mental models of the machine learning system changes during the workshop and that students initially hold a variety of conceptions that can be structured along two facets (“internal model” and “learning process”). Some of these conceptions are useful for learning, while others may present an obstacle - an information that is relevant for designing teaching about machine learning in the context of general education.