Teaching AI models to express uncertainty in words — calibrated hedging in responses

AI Impact Summary

The change centers on enabling AI models to surface explicit uncertainty in their outputs, using hedges or probability-language rather than definitive statements. This supports risk-aware decision-making and can improve user trust by aligning responses with evidence strength. To realize value, teams will need to update UI and downstream analytics to properly present and interpret uncertainty signals, and establish calibration standards to avoid inconsistent hedging across tasks.