Focus
An LLM agent–based search and recommendation system for real-world experiences. ExperienceBot treats experience discovery — not document retrieval — as the core search problem, and uses an agentic loop to elicit, surface, and reason over the latent dimensions of an experience a user actually cares about.
About
ExperienceBot is an LLM-powered agent for helping people with realizing desired experiences out in the world (for example, finding places suitable for a long, meaningful conversation). Currently, LLMs have the appearance of being able to respond meaningfully to seemingly any queries about experiences and to provide meaningful recommendations. But in practice, many LLM recommendations can miss hidden facets of experiences that a person cares about, but who may also struggle to state the various latent dimensions of the experience that matter to them.
Through considering and surfacing experience conceptions, user perspectives, and practical issues, ExperienceBot aims to provide LLMs with a deeper experiential understanding of human activities and to better facilitate people using LLMs to conduct rich queries and explorations into experiences that may be of interest to them. This includes responding to people’s latent experience conceptions, and supporting them in reconceptualizing when certain experience conceptions are unlikely to be realizable in their settings or environments. We may also explore how to help people understand what the experiences they crave are telling them about themselves.
Affiliation
This research is conducted at Northwestern’s Design, Technology & Research (DTR) Lab.