Access & Terms of Use
embargoed access
Embargoed until 2024-09-26
Copyright: Chong, Terrence
Embargoed until 2024-09-26
Copyright: Chong, Terrence
Altmetric
Abstract
As chatbots are deployed for service interactions with customers on the organizational frontline, they are designed to be humanlike (anthropomorphized) to increase acceptance and usage. Although the use of chatbots continues to rise, there is a lack of guidelines on effective chatbot design and its impact on customer responses. Drawing from social cognitive theory (SCT) and stereotype content model (SCM), this thesis consists of three essays that theoretically propose and empirically test the underlying mechanisms of the effect of embodied conversational agent (ECA, a form of chatbot with digital face/head or body) design on customer outcomes.
Essay 1 (Chapter 2) conceptualizes three aspects of ECA design (i.e., anthropomorphic role, appearance and interactivity) and presents a framework on how they impact service outcomes by influencing people's agency.
Essay 2 (Chapter 3) builds on Essay 1 and introduces ‘value-by-proxy’ as a two-stage process that accounts for how the three anthropomorphic affordances relevant to ECAs (i.e., anthropomorphic role, visual appearance and conversational style) translate into relevant customer outcomes in the delivery of online financial coaching service. The essay uses an experimental approach that simulates an interactive service encounter with an ECA and incorporates implementation intentions with regard to (non)use and advice compliance, which is integrated with a Gabor Granger study to assess the willingness to pay. Based on a sample of 596 US-based respondents recruited from a Prolific panel, the results provide evidence on the serial mediation of the impact of anthropomorphic affordances on customer outcomes.
Essay 3 (Chapter 4) conceptually and empirically explores how the dimensions of warmth and competence are afforded through prosody (enthusiastic vs. calm voice) and interaction styles (socioemotional vs. task-focused) in establishing perceptions of virtual rapport and in turn customer usage and compliance intentions. Data were collected from two studies: 390 US-based respondents recruited from a Prolific panel participated in an online experiment in Study 1, and 212 US-based respondents from a Conjoint.ly panel participated in a choice-based conjoint experiment in Study 2. The findings demonstrate a process of moderated serial mediation such that feelings of rapport are driven by believability and that this relationship is enhanced by a willingness to suspend disbelief.
