Experiential Computing Platform

This project develops an experiential computing platform that helps designers understand how the same activity plays out very differently across cities and communities. Instead of treating venues as generic categories like “bakery” or “cafe”, our system analyzes millions of user reviews to uncover the local rituals, expectations, and constraints that actually shape experiences in those places. It then uses a conversational interface where designers can describe a target experience, such as a quiet, low-cost first date in Chicago, and see where that experience fits or fails in specific neighborhoods.

At the core is a multi-agent LLM architecture that separates high-level conception from on-the-ground realization. Specialized agents query a backend that returns detailed, step-by-step information about what the local environment can support, along with review snippets that reveal what people value or dislike. This design forces the LLM to ground its critiques in evidence rather than hallucinating plausible stories, building on recent work on LLM hallucination and retrieval augmented generation.

The result is a practical tool that surfaces where universal design assumptions break for real communities and helps teams redesign experiences in ways that respect local realities.