This work introduces a task to forecast long-term skill development in conversationalists, specifically in mental-health crisis counselors. It focuses on identifying moments where counselors struggle and analyzing how they adapt responses over time.
The approach captures initial difficulties and tracks how responses evolve when similar situations are encountered later. It learns which early adaptations are indicative of future improvement, even months or years ahead.
This method outperforms baseline models that learn directly from conversation transcripts, providing a way to prioritize support for counselors who need it most. It offers insights into skill development that can inform training and supervision strategies.
Understanding these predictive signals can help improve mental-health support systems and optimize resource allocation for counselor training.
Source: https://arxiv.org/abs/2609.04350