Enhancing AI Interaction through Co-Construction: A Multi-Faceted Workshop Framework

The boom of research topics such as AI literacy and explainable AI (XAI) intensified political discussions about the right for meaningful explanations of the logic involved leave no doubt about the necessity of AI education. However, technologies that we find in our every lives often hide architectural aspects to make their benefits accessible for everyone. Additionally, performance seems to trade-off for explainability, leading to a black-box system not understandable for users and sometimes even experts. To address this issue this paper introduces a workshop framework designed to enhance interactions with large language models (LLMs) - in particular ChatGPT. With emphasis on observation, analysis and hands-on interaction, the workshop framework pursues three different goals: (1) as an event for scientific research, (2) as a learning event and (3) as science communication work to integrate the public more closely into research. Pre-post-test data suggests a shift in participants' interaction patterns when using ChatGPT. Participants not only questioned the AI's responses but also reflected on their own understanding, asking follow-up questions or offering suggestions on how to better explain concepts to the AI.