Machine learning role playing game: Instructional design of AI education for age-appropriate in K-12 and beyond

Abstract

The study on K-12 artificial intelligence (AI) education poses three challenges: ''(i) What topics in the machine learning (ML) process should be hidden from K-12 students?” '' (ii) How do you teach K-12 students the mathematical model behind the ML process?” and '' (iii) How does AI education influence the acceptance of AI technology?”. This paper addresses challenges (i) to (iii). We developed a machine learning role-playing game (ML-RPG) and had 166 participants from lower grades of elementary school to the elderly use it. Participants used ML-RPG to role-play the ML process of acquiring data, representing data with graphs, inferencing based on if-then rules, and optimizing parameters based on loss functions in the ML model. Challenge (i) was assessed using an achievement test. The results showed that the ML process of perception at K1-3, data representation and reasoning based on if-then rules at K4-6, and simple ML model learning and evaluation at K7+ can be unveiled. Challenge (ii), K7+ were able to understand the optimization process of the ML model by role-playing to minimize the loss function shown on the graph. Challenge (iii), it was shown that most people who feel vague anxiety and fear about AI change to a positive impression of AI due to the following three factors: (a) participants understanding the decision criteria of AI, (b) increased self-efficacy among participants, and (c) participants realizing that AI can be applied to familiar things.

Keywords

• Data science applications in education
• Games
• Early years education
• Secondary education
• Post-secondary education