Deal, or no deal (or who knows)? Forecasting Uncertainty in Conversations using Large Language Models

Abstract: Effective interlocutors account for the uncertain goals, beliefs, and emotions of others. But even the best human conversationalist cannot perfectly anticipate the trajectory of a dialogue. How well can LLMs represent inherent uncertainty in conversations? We propose FortUne Dial, an expansion of the long-standing "conversation forecasting" task: instead of just accuracy, evaluation is conducted with uncertainty-aware metrics, effectively enabling abstention on individual instances. We study two ways in which LLMs potentially represent outcome uncertainty (internally, using scores and directly, using tokens) and propose fine-tuning strategies to improve calibration of both representations. Experiments on eight difficult negotiation corpora demonstrate that our proposed fine-tuning strategies (a traditional supervision strategy and an off-policy reinforcement learning strategy) can calibrate smaller open-source models to compete with pre-trained models 10x their size.