Relationship between trust and entrainment in speech
Abstract
We have designed experimental scenarios and collected data from Slovak and Argentinian Spanish testing the effect of manipulating acoustic-prosodic features of synthesized voices, to either entrain or disentrain with respect to the human user's voice, on trust of users toward the avatars with these synthesized voices. The results of these experiments and our continued exploration of human-human entrainment in spoken dialogue corpora provide a rather complex and multidimensional picture of prosodic entrainment and its potential for calibrating trust in human-machine spoken interactions. The key results include: 1) not only entrainment but also disentrainment in terms of acoustic-prosodic features might facilitate trust of humans towards computer avatars, 2) different prosodic features affect the trust-entrainment relationship differently, 3) most of our results were consistent for both Slovak and Argentinian Spanish, which points to complex, yet systematic and cross-culturally valid, nature of the trust-entrainment relationship, and 4) the observations from human-human dialogues are generally consistent with human-machine experiments and offer also novel factors, such as biological sex and task-role, that affect speech entrainment and the entrainment-trust relationship.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 25, 2018
- Accession Number
- AD1061195
Entities
People
- Štefan Beňuš
Organizations
- University of Constantine the Philosopher