Design and Implementation of Conversational User Interfaces for Health

Abstract

Conversational user interfaces (CUIs) have the potential to support users across varied health domain areas. Yet barriers remain to the implementation and adoption of CUIs, such as lack of trustworthiness and consideration for cultural context. There exist a number of conversational-based interventions that have been proven to be usable and effective in addressing health conditions. However, there are still few conversational-based health interventions implemented and studied in real-world settings. My dissertation research examines the design and implementation of CUIs for health interventions, where I integrate human-centered design (HCD) and implementation science methods to understand stakeholder needs and optimize implementation strategies. Drawing from methods and frameworks in HCD and implementation science, I explore CUI design considerations and challenges in two health contexts: 1) social needs screening in a large public hospital emergency department, and 2) breast cancer screening outreach for Black/African American women. First, I adapt human-centered design approaches to improve the design of CUIs by developing and validating heuristics for conversational agents. I conduct this research to demonstrate that conversational user interfaces require unique design guidelines and considerations. Second, I deploy a conversational user interface in a real-world context to evaluate its effectiveness in supporting patient engagement. Through such deployment, I highlight implementation challenges to integrating a social needs screening chatbot into an emergency care setting, as well as individual, contextual, and intervention-related factors that may influence engagement in CUI interventions. Lastly, I engage in multidisciplinary design work to integrate methods from HCD and implementation science to improve CUI design for health interventions. I describe the development of a chatbot intervention aimed to facilitate breast cancer screening outreach. In summary, my dissertation demonstrates that adaptation of established usability heuristics for conversational user interfaces can lead to improved usability and engagement. I also discuss how an integrated approach of human-centered design and implementation science methods may combine the strengths of both disciplines in the design of chatbot implementation strategies. My dissertation makes (1) methodological contributions through the development of usability heuristics, (2) artifact contributions through the development of CUIs and through the design recommendations that arise through real-world deployments, and (3) empirical contributions through studying how CUI design components relate to engagement.