Prediction of trust in scripted dialogs using neuro-fuzzy method

Design of effective communication dialogs can help to determine trust between humans. This paper reports an experimental study to explore and predict trust using two different scripted dialogs. The first dialog represented a business scenario, while the second was a fire rescue. Trust was measured using subjective psychological criteria of Ability, Benevolence and Integrity which were embedded in the dialog. These were correlated with objective physiological features of human facial expressions, voiced speech and camera-based heart rate. Using neuro-fuzzy method, the subjective trust data was analyzed separately. The results showed that subjective measures of Benevolence and Integrity for fire rescue scenario could predict trust better than the business scenario. The power of prediction was almost identical for the Ability measure. The findings suggest that trust is associated with the type of scenario, and that both subjective and objective measures are important in predicting trust.