Representing discussions: re-fract

Multi-party negotiations, especially related to complex landscapes with varied land-uses and stakeholders, involve a dense web of spoken interactions, social tensions, and shifting intents. Interpreting and communicating the outputs of such discussions can be challenging, with a diversity of perspectives, agreements, and conflicts to summarise.

The re-fract project introduces a visual analytical toolkit designed to transform live recordings into a clear, interactive map of the discourse.

Moving beyond simple transcription, the system utilises advanced AI to perform real-time Semantic Understanding and Theory of Mind modelling. This allows the toolkit to reconstruct and track the complex social dynamics of the room, thus identifying not just what is being said, but how participants are engaging with one another’s ideas. From a Human-Computer Interaction perspective, our focus is on reducing the cognitive load of the mediator and participants by visualising these implicit patterns of influence, sentiment, and stance.

By providing an intuitive, data-driven grounding for discussion sessions, the system empowers stakeholders to reflect on their negotiations and conversations - coming away with not just the concluding remarks, but an understanding of the entire process. Our aim is for such transparency to better represent discussions, foster mutual understanding, and facilitate more equitable and sustainable outcomes for all involved.