LuminAI
LuminAI is an exhibit concept that utilizes creative, embodied interactions to engage users in learning about agent reasoning and decision-making processes. LuminAI builds on an existing AI installation in which participants can improvise movement together with an AI dance partner. The dance partner senses participant movements via a motion capture sensor and responds with movements in its memory that it deems to be similar. This design workbook sketch proposes expanding LuminAI into an educational exhibit by allowing learners to 1) choose between different datasets for the dancer (e.g. hip-hop dance vs. ballet); 2) visually inspect the contents of the datasets using a VR-based visualization tool; and 3) dance with an AI trained on the dataset they selected. These options would not only maek the agent reasoning and knowledge representations more transparent, but it would also allow learneres to engage in concepts such as training dataset bias and machine learning model refinement on a creative and embodied way. Situating an AI learning experience in the domain of dance would additionally allow learners to explore different modes of AI-human interaction.
Competencies: ML steps; representations; decision-making; learning from data; interdisciplinary
Design Considerations: explainability; new perspectives; social interaction; embodied interaction; opportunities to program; leveraging learners’ interests