Event Actions
Mining Multimodal and Multidimensional Patterns of Trustworthy Behavior
Abstract:
Interpersonal trust takes many forms and can be influenced by small, challenging-to-quantify events. However, no study investigates relationships between multimodal behavioral patterns and different dimensions of trust. We conducted a study involving 500 participants to explore the connection between an individual’s behavior and how trustworthy they are perceived by others. Participants were shown videos of people being interviewed and were asked to fill out a questionnaire rating the trustworthiness of the person in the video. We developed a novel method to extract common behavioral patterns linked to different levels of trustworthiness, and our analysis revealed that these associations are highly variable. While some behavioral patterns appear to be commonly linked to particular dimensions of trust, they do not provide clear relationships. Conversely, less common behaviors are more conclusive. We also discovered that behavioral patterns differ depending on the type of trust. These findings could be useful for developing machines that can evaluate the trustworthiness of their users.
Committee:
- Seongkook Heo, Committee Chair, (CS/SEAS/UVA)
- Afsaneh Doryab, Advisor (SYS, CS/SEAS/UVA )
- Tariq Iqbal (SYS, CS/SEAS/UVA )
- Jack Davidson (CS/SEAS/UVA)