Machine Learning for Non-Major Data Science Students: A White Box Approach

Data science is a new field of research that has attracted growing interest in recent years as it focuses on turning raw data into understanding, insight, knowledge, and value. New data science education programs, which are being launched at an increasing rate, are designed for multiple education levels and populations. Machine learning (ML) is an essential element of data science that requires an extensive background in mathematics. Whereas it is possible to teach the principles of ML only as a black box, novice learners might find it difficult to improve an algorithm's performance without a white box understanding of the underlying ML algorithms. In this paper, we suggest a pedagogical method, based on hands-on pen-and-paper tasks, to support white box understanding of ML algorithms for learners who lack the level of mathematics knowledge required for this purpose. Data were collected using a comprehension questionnaire and analyzed according to the process-object theory borrowed from mathematics education research. We present evidence of the effectiveness of this method based on data collected in an introduction-level data science course for graduate psychology students. This population had extensive psychology domain knowledge, as well as an established background in statistics, but had gaps in mathematical and computer science knowledge compared with data science majors. The research contribution is both practical and theoretical. Practically, we present a learning module that supports non-major data science students' white box understanding of ML. Theoretically, we propose a data analysis method to evaluate students' conceptions of ML algorithms.