TLDR: A UCLA production of the musical Xanadu used AI to create an immersive, interactive experience for over 500 audience members. By framing the show as a modern ritual, AI transformed audience sketches and dance moves into virtual scenery and choreography, fostering collaborative human-AI engagement and addressing the limitations of single-user AI design paradigms. Performers guided the audience through these interactions, creating a reciprocal feedback loop where human actions influenced AI-generated content, and AI, in turn, guided human participation.
A recent project at UCLA explored a novel way to integrate artificial intelligence into group experiences, moving beyond the common single-user interactions we see with most AI applications. Researchers and artists collaborated on an immersive, interactive production of the roller disco musical Xanadu, using the concept of ritual as a framework for human-AI engagement.
The production, staged in Spring 2025, involved over five hundred audience members who actively contributed to the show. Instead of just watching, attendees were invited to participate by creating sketches and performing jazzercise moves. These human inputs were then translated by vision-language AI models into virtual scenery elements and choreographic prompts, visibly reshaping the virtual world of the musical.
The core idea behind this innovative approach was to view the theatrical experience as a modern-day ritual. In this framework, audience contributions were seen as ‘offerings’ that AI transformed into components of the ritual. Performers acted as ‘ritual guides,’ demonstrating how to interact with the technology and organizing audience members into groups. The AI systems themselves were conceptualized as ‘instruments’ played by humans, where sensing, generative components, and stagecraft worked together to create systems that could be mastered over time.
A key aspect of this design was the ‘reciprocity of interaction.’ This meant that the show’s AI machinery not only responded to human input but also guided human behavior, completing a feedback loop. For example, audience sketches were transformed into polished images or 3D assets that matched the show’s unique vaporwave aesthetic. Performers would model the drawing process, encouraging collective contribution over individual perfection, which helped reduce audience anxiety and fostered collaboration.
In one particularly engaging segment, an AI component guided audience movement. An actor, playing the Oracle, consulted an iPad connected to a vision-language model. This AI analyzed the audience’s previous sketch-driven offerings and generated a short poem, which then described three jazzercise-style moves. The Oracle and the Muse of Dance taught these moves to the audience, who then performed the choreography. Depth cameras tracked the collective motion, and if the crowd’s energy and alignment met a certain threshold, the virtual environment would respond dynamically with lighting shifts and scene updates. This synchronized movement was found to enhance trust and group affiliation among participants.
The project aimed to address a significant gap in current AI design paradigms, which predominantly focus on individual user interfaces. By framing the interaction as a ritual, the team successfully integrated linear narrative, character identity, music, and audience participation, exploring how AI can support group creativity and play. This approach draws parallels to ancient Greek rituals, where lay participants were guided by experts to create offerings for non-human entities, a structure mirrored in the Xanadu production.
Also Read:
- Experiential AI Learning: Instructors Share Insights on Engaging Non-STEM Students with Real-World Scenarios
- Bridging the AI Literacy Gap: How Community College Instructors Envision Teaching AI to All Students
While the project demonstrated the potential for AI in collective experiences, the authors acknowledge limitations, including the bespoke nature of the production and the informal assessment of audience experience. However, the work offers valuable insights into designing AI systems for collaborative, multi-user engagement, suggesting future applications in areas like collaborative fandom and collective meaning-making. You can read more about this fascinating research at the full paper.
