Let's Go for a Drive: Exploring AI's Societal Impact in K-8 Education with an Interactive Self-Driving Car Tool
Pranathi Rayavaram, Sashank Narain, Fred G. Martin · 2024
This innovative practice full paper discusses the development of an interactive tool designed to educate middle school students on the ethical considerations and societal impacts of artificial intelligence (AI). As AI technologies become more embedded in our daily lives, the younger generation must grasp the implications of algorithmic bias and its societal effects. Our tool aims to deepen this understanding by focusing on self-driving cars-a relevant and significant example of AI technology. The interactive tool incorporates three advanced image recognition models trained on diverse datasets, including traffic cones, animals, and pedestrians. Through the tool's interface, students can choose one of these models to apply in a self-driving car simulation and select different obstacles for the car to encounter, such as traffic cones, animals, and pedestrians. This hands-on simulation highlights the importance of comprehensive AI model training, showcasing how well-trained models help avoid collisions and the risks associated with encountering untrained obstacles. It engages students by demonstrating how developers' AI training decisions can significantly influence end-user experiences. Moreover, the tool emphasizes the need for diverse and representative data in building fair and robust AI systems. To assess the effectiveness of this educational tool, we conducted a two-day AI exhibit attended by 26 middle-school students from grades six to eight. The effectiveness was evaluated through posttrial questionnaires to measure the students' understanding of several key concepts: the development of resilient AI models, the societal impacts of AI, and the ethical considerations of road safety in the context of AI. The results showed that 84.6 % of the participants understood how poor training decisions could impact AI outcomes, and about 96 % recognized the necessity for diverse data.