From Demo to Design in Teaching Machine Learning

Karl-Emil Kjær Bilstrup, Magnus Høholt Kaspersen, Ira Assent, Simon Enni, Marianne Graves Petersen

Abstract: The prevalence of artificial intelligence (AI) and machine learning (ML) technologies in digital ecosystems has led to a push for AI literacy, giving everybody, including K-12 students, the necessary knowledge and abilities to engage critically with these new technologies. While there is an increasing focus on designing tools and activities for teaching machine learning, most tools sidestep engaging with the complexity and trade-offs inherent in the design of ML models in favor of demonstrating the power and functionality of the technology. In this paper, we investigate how a design perspective can inform the design of educational tools and activities for teaching machine learning. Through a literature review, we identify 34 tools and activities for teaching ML, and using a design perspective on ML system development, we examine strengths and limitations in how they engage students in the complex design considerations linked to the different components of machine learners. Based on this work, we suggest directions for furthering AI literacy through adopting a design approach in teaching ML.