Qingxiao Zheng's Dissertation Proposal Defense
PhD student Qingxiao Zheng will defend their dissertation proposal: Ethical Design and Democratization of Conversational AI: Enhancing User eXperience through Large Language Models.
Abstract:
Conversational AI is considered highly promising and has seen growing applications across various domains. This dissertation explores the design and evaluation of conversational AI (CA) as a facilitator in human-to-human communication, focusing on user experience (UX) evaluation of CAs, ethical design considerations, and democratizing CA technology. The contributions of this dissertation are three-fold.
First, I conduct a comprehensive literature review of conversational AI research and introduce a user experience (UX) evaluation framework that enables designers to assess users’ perceptions and interactions with CAs in diverse contexts, such as communication, engagement, connection, and relationship maintenance.
Second, I illustrate ethical CA design that takes into account social boundaries, including aspects such as privacy, disclosure, and identification. Specifically, I demonstrate the impact of CA in facilitating emotional communication between peers, such as long-distance couples. I showcase that an ethical CA can be designed to improve couples’ conversations by introducing humor and fostering deeper connections across different stages of their relationship.
Third, I explore new approaches to demoncraticizing CAs for multi-stakeholder engagement. In particular, I develop unique interaction designs that enable service providers to create conversational agents, which can be employed to augment their service responses to requesters. Our approach advances the existing participatory design practices by incorporating a principle of “Beginning with the End in Mind.” Namely, throughout the design stage, users consider the evaluation metrics of the CAs and execute different strategies to improve the UX by utilizing large language models (LLMs), such as offering emotional support or delivering concise responses.
Although this study will be conducted in the context of designing a CA for librarians to serve patron users, this dissertation’s contributions, including the proposed UX evaluation framework, ethical design suggestions, and democratization strategies empowered by LLMs, have both theoretical and practical implications in broader application domains.
Committee members include Associate Professor Yun Huang (Chair); Dr. Vera Liao, Microsoft AI; Associate Professor Yang Wang; and Professor Mike Yao.
Questions? Contact Qingxiao Zheng.