Leveraging Conversational User Interfaces and Digital Humans to Provide an Accessible and Supportive User Experience on an Ophthalmology Service

Abstract

Designing E-Health services that are accessible, engaging, and provide valuable information to patients is an endeavor that requires research and validation with potential users. The information needs to be perceived as trustworthy and reliable, in order to promote people’s ability to make informed decisions about their health. This Master’s thesis work focused on understanding the potential of conversational user interfaces (CUIs) featuring digital humans (DHs) as communication agents to provide healthcare-related information to users. The case study underlying the research was proposed by Roche: the company wanted to create an informational ophthalmology website featuring a digital human to substitute the traditional text-based website. The main goal of this work was to understand whether CUIs and DHs can provide a higher level of accessibility and engagement for users, with a special focus on people starting to live with low vision (potential ophthalmology patients).

Managing to address these aspects would allow providing a better user experience for people visiting the website. Since digital humans are not yet extensively adopted in the healthcare domain, few design guidelines are available. The work employed a human-centered design approach, to gather requirements and feedback from users, and led to defining six guidelines and an extensive set of observations about user experience and accessibility. These guidelines are: ensure that the digital human is as realistic as possible; create a clear and easy to follow conversation; present options simply and allow flexibility in choice methods; provide a text version of the content; ensure that easy and self-explanatory navigation; ensure compatibility with assistive technologies and provide flexibility, personalization and integration.

The user research was divided into two phases. First, an exploratory research session was conducted, where ten participants were recruited to investigate the needs and expectations of people living with eye conditions towards an informative service and their first impressions of DHs. This session employed the semi-structured interview methodology, and the results informed the further development of the service. When the first proof of concept prototype version of the website was built, an evaluative research phase with eighteen participants was conducted. This session was conducted using the participant observation methodology paired with semi-structured elicitation interviews. Afterwards focus group sessions were organized to have the participants further discuss their experience. The user-based research was paired with expert evaluation using the cognitive walkthrough methodology and a simplified WCAG 2.1 accessibility assessment. Combining the two approaches gave a good overview of the merits and issues of the approach.

The results of the research allowed building a good understanding of the positive and negative aspects of using a digital human as an agent in a conversational user interface. Users generally appreciated the concept: they found it engaging, trustworthy and easy to use. However, there are some aspects that could not be addressed during this research, and which need further understanding. The primary areas that need to be addressed are guidance, navigation, and error management. Nonetheless, the positive feedback gathered from the participants of the evaluation sessions proves that it is worth investing in the research and development of this relatively new kind of service. In fact, the results of the work show that having a digital human as agent for a conversation-based informative service in healthcare has strong potential, in terms of both accessibility and engagement.