Abstract
dc:description.abstractDriver anger degrades vehicle control and safety, and in-vehicle conversational agents are increasingly proposed as a means of supporting drivers through emotionally difficult moments. Empathy is commonly treated as a desirable property of such agents, yet an agent has no emotional capacity in the sense the construct presupposes: what it has is a repertoire of utterances. This dissertation therefore operationalizes empathy as a verbal design strategy and asks which realization of it supports drivers, and whether the version that is best received is also the version that works best. Study 1 (N = 60) compared cognitive empathy, which names the driver's likely state and explains why the situation warrants it, against affective empathy, which expresses a parallel reaction of the agent's own, and against navigation-only speech, under a pre-drive anger induction. Study 2 (N = 48) then crossed the same three empathy conditions with agent embodiment, comparing a physically present humanoid robot against a voice-only agent. Study 3 (N = 48) crossed empathy condition with conversational interactivity, comparing a pre-scripted agent against one powered by a large language model while holding the robot, the voice, the timing, and the empathic utterances constant. All three studies measured self-reported affect, physiological response, perceptions of the agent, and vehicle control. The findings converge on a separation between how an agent is received and what it does. Cognitive empathy was associated with the most controlled vehicle handling in Study 1 and with the highest rate of active hazard response in Study 3, and it incurred far less of the penalty in perceived safety that emotionally expressive speech carried in Studies 2 and 3. Affective empathy, by contrast, was rated warmest, most animate, and most socially present, and was associated with faster cardiac settling in Study 1. However, it did not outperform cognitive empathy on any measure of vehicle control, and it reduced how safe the agent felt to be with in Studies 2 and 3. Embodiment lowered every dimension of perceived workload largely without changing how empathic content was evaluated. Conversational interactivity produced the strongest perceptual effects observed in this dissertation, nearly doubling perceived social presence, but these gains did not carry over: conversational capability had no effect on vehicle control and influenced physiological response only at the peak of the cardiac reaction. When drivers could converse freely with the agent, they asked overwhelmingly about information and company, with expressions of emotion accounting for a small fraction of what they said. Across the three studies, drivers' evaluations of the agent did not move together with their physiological responses or their driving behavior: ratings did not predict physiological response at the individual level, and the design that drivers judged warmest was not the design associated with the most controlled driving. On this basis, the dissertation argues that empathic in-vehicle agents should be evaluated on driver state and behavior rather than on social impression alone, that cognitive empathy offers the most defensible default for safety-critical settings, and that embodiment and conversational capabilities may contribute more to user acceptance than to physiological or behavioral regulation.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Industrial and Systems Engineering
- Department dc:contributor.department
- Industrial and Systems Engineering
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Choe, Mungyeong
- Chair dc:contributor.committeechair
-
- Jeon, Myounghoon
- Committee members dc:contributor.committeemember
-
- Lau, Nathan Ka Ching
- Alsaid, Areen
- Klauer, Sheila G.
Subjects
dc:subject × 6Rights
dc:rights- Statement dc:rights
-
- In Copyright
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:47851