Paper ID

3378

Paper Type

full

Description

The utilisation of embodied conversational agents (ECAs) to build a human-agent working alliance holds promise to promote health behavior change and improve health outcomes. Although ECAs have been shown to build empathic relationships with users, there is no complete framework to model working alliance. In this paper, we developed a framework that is grounded on theories and findings from social science and artificial intelligence to design a cognitive architecture for a user-aware explainable ECA. An empirical evaluation with 68 undergraduate students found differences in the efficacy of explanation to change behavior intention, build trust and working alliance depending on gender, stress levels and achievement aims; confirming the imperative of incorporating shared planning and user-tailored explanation in one framework. The empirical evaluation was limited in tailoring the explanation to the user’s beliefs only; however, the analyses confirmed the need for considering adequate user information such as user’s goals and preferences to build a user-aware explainable agent for behavior change towards improved health outcomes.

Share

COinS
 

Modelling Working Alliance Using User- aware Explainable Embodied Conversational Agent for Behaviour Change: Framework and Empirical Evaluation

The utilisation of embodied conversational agents (ECAs) to build a human-agent working alliance holds promise to promote health behavior change and improve health outcomes. Although ECAs have been shown to build empathic relationships with users, there is no complete framework to model working alliance. In this paper, we developed a framework that is grounded on theories and findings from social science and artificial intelligence to design a cognitive architecture for a user-aware explainable ECA. An empirical evaluation with 68 undergraduate students found differences in the efficacy of explanation to change behavior intention, build trust and working alliance depending on gender, stress levels and achievement aims; confirming the imperative of incorporating shared planning and user-tailored explanation in one framework. The empirical evaluation was limited in tailoring the explanation to the user’s beliefs only; however, the analyses confirmed the need for considering adequate user information such as user’s goals and preferences to build a user-aware explainable agent for behavior change towards improved health outcomes.