A workshop on artificial intelligence biases and its effect on high school students’ perceptions

Abstract

This paper introduces a workshop aimed at concurrently addressing technical concepts and ethical considerations on artificial intelligence (AI) biases, with an emphasis on societal and automation biases. Unlike conventional approaches that often prioritize either the technical intricacies or the ethical implications of AI, our workshop integrates these dimensions in parallel. Through a series of activities, we illustrate how design decisions made by individuals involved in AI development, such as defining classes and selecting training sets, can introduce biases into AI models. We also explore errors or biases in model decisions, shedding light on the nuanced challenges of AI development. The workshop’s impact on high school students’ perceptions of AI technology was assessed through pre and post-tests. Statistical analysis revealed a significant reduction in students’ agreement with statements regarding the absence of AI societal biases, the lack of influence of AI designers on AI behavior, and the superiority of AI solutions over human alternatives. While perceived as a highly positive and engaging experience, the workshop was also recognized as a practical and motivating endeavor, aligning with our didactic approach emphasizing experiential learning over theoretical exposition.