Tag
Proposes a decision mechanism for forecasting conversational derailment that decouples trigger decisions from derailment likelihood estimation, using forward-looking simulations to defer alerts when recovery is plausible, reducing false positives.