Algorithm Selection to Identify Brain Dominance Through Game-Based Learning. An Ethical Perspective

Abstract

Artificial Intelligence (AI) is ubiquitous employed in homes, workplaces, and schools. Concerns have been raised about the possible harmful impact of AI including the usurpation of trust and freedom. One potential way to combat mistrust is by providing algorithmic transparency alongside a high performing model. The aim of this paper is to describe the process used to ethically select an AI algorithm to be used on a digital game designed to identify the preferred learning style of children while learning computer programming. 5 models were compared based on explainability, bias, nondiscrimination, fairness, and performance before selecting the most ethically sound algorithms. It was concluded that Decision Tree and Random Forrest models perform best in both transparency and performance outperforming other, black-box models in both accuracy and explainability at identifying users preferred learning style.

Ähnliche Arbeiten