Exploring data visualizations as a tool to make ai more accessible for qualitative researchers in the healthcare field
More details
Hide details
University of Waterloo Canada
Publication date: 2023-04-26
Popul. Med. 2023;5(Supplement):A625
Background and objective:
Qualitative data analysis allows healthcare researchers to understand lived-in experiences of patients and practitioners. However, it requires time and effort, which is not always available. A potential way to assist is to use artificial intelligence as a tool to help researchers with data exploration. However, many qualitative researchers do not have programming skills to use AI and are reluctant to lose their sense of agency and serendipity when conducting research. As a potential way to bridge this gap, we explored the use of data visualizations to foster researcher agency and make AI more accessible.

We used Design Science Research to develop a series of prototypes. First qualitative semi-structured interviews were conducted with 5 qualitative researchers to gage how they interacted with their data and what they expected of data visualization for working with AI. The interviews were guided by the concept of agency and how it can be fostered through media interaction. Then, the results were compiled into guidelines for the development of 4 datavis prototypes, each focusing on fostering agency through a different channel.

Researchers reported needing freedom to move within a visualization the most. High levels of tactile interactivity increased the trust in the AI and made participants feel like they were working as a partnership. Additionally, freedom to co-create the visualization instead of being presented with a finished product was cited as satisfying. Both elements made participants feel ownerships over their own research and a sense of camaraderie with the AI.

For qualitative researchers, agency in data analysis is heavily correlated with freedom of choice and movement. Those elements directly impact their sense of ownership of data and trust in AI. Overall, data visualizations seem to show promise in making AI more accessible for people with no programming experience.