#654: Learning through Chatbot Design: Examining Children’s AI Understanding and Hallucination Awareness

Young learners increasingly interact with AI chatbots in classrooms and daily life, as educators explore their potential to support learning. While chatbots can personalize instruction and enhance engagement, they also produce hallucinations, which are plausible but incorrect responses. Because such errors arise from the probabilistic nature of generative AI, it is essential to help students develop literacy in recognizing and addressing them. Grounded in constructionist approaches to AI literacy, this study engaged 47 middle school students in designing and testing their own chatbots using a LUMI environment equipped with hallucination detection tools. Through hands-on creation and peer evaluation, students developed a deeper understanding of how AI systems function and where their limitations emerge. Findings show significant gains in AI understanding, hallucination awareness, and students’ ability to propose concrete detection and mitigation strategies. The study demonstrates how we can cultivate critical and reflective engagement with youth through emerging AI technologies.