| Summary: | Computer agents are no longer considered as the interface merely the instrument or medium of human interaction, but more importantly, the partner engaging in communicative exchanges (Gunkel, 2012). While a vast group of studies has investigated the design addressing both task-solving skills and social capabilities (such as self-disclosure), the extent to which the mechanism of human-computer trust and relationship development has remained inconclusive. This project presents the results of a 2 (ECA communicates self-disclosure vs. non-self-disclosure) × 2 (task with high objectivity vs. low objectivity) online experiment (N = 200). Consistent with existing algorithm aversion research, this project found that people may fundamentally perceive embodied conversational agents (ECAs) as incompetent for performing subjective (vs. objective) tasks and reluctant to interact with them. ECAs’ humanlike communication style, e.g., self-disclosure, showed positive impact on warmth, which in turn, facilitated trust and relationship development. When an ECA performed high-objective tasks, participants rated it to be warmer when is incorporates self-disclosure during verbal communication, which in turn, promoted trust and relationship development.
|
|---|