Browsing by Author "Papandreou, Amanda"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    What Kind of AI Users Are There?
    (2024-04-26) Jones-VanMieghem, Cassandra; Papandreou, Amanda; Dolan, Levi
    INTRODUCTION • The field of Human-Computer Interaction (HCI) has long used the technique of personas in theory and practice to enhance human-centered design (Chang et al., 2008). • Generative AI (GenAI) introduces new HCI issues due to inherent generative variability (Muller et al., 2023) and shifts in the meaning of control (Geyer et al., 2024). • The effects of this variability are part of the ongoing discussion about the role of emotion in HCI (Wadley et al., 2022), and are being felt across various domains such as education, biomedical research, and administrative tasks. • Persona development in GenAI HCI is a design technique that can be adapted to student and professional development to potentially improve learner and user experience. OBJECTIVES • To identify characteristics of GenAI interactions in library consultations, instruction sessions, and trainings. • To propose an initial theoretical HCI persona framework for AI users, focused on users of health systems-related AI applications. METHODS • We collectively wrote reflections on the utilization of AI and attitudes of the medical students, staff, and faculty that we encountered on the job from November 2023-January 2024 . • We attended various AI professional development sessions from November 2023-April 2024 , during which we took notes regarding the types of questions posed by attendees to capture prevalent concerns and interests. • We discussed our consolidated observations and identified major attitudinal themes. • We developed a Four-Persona Framework to categorize these themes. • We searched for points of contact with these themes in recent HCI and medical literature as an initial exploration of the framework’s generalizability. RESULTS Four Persona Framework • Unconscious User (Don’t know/Don’t care) o Survey of general US public found that they believe humans make better decisions than AI (5 studies n=<1000; Dietvorst et al., 2014). o Survey of emergency and trauma surgeons worldwide found that 73.3% could not define a list of AI-related terms (n=200; De Simone et al., 2022). o Survey of radiation oncologists and medical students in China found that they were favorable toward using AI in healthcare systems due to reduced workloads, unlike the public’s centering of risk/intent in usage choice (Zhai et al., 2021) • Avoidant User (Dangerous) o Survey takers from the general US public who indicated a low level of mistrust of human decision makers mistrusted AI more than humans (n=187; Lee & Rich, 2021) o Survey of physicians in Germany found that 48.2% agreed that using AI prevents doctors from learning how to correctly assess a patient (n=294; Maassen et al., 2021) o Survey of medical and dental students in 63 countries found that 43.2% disagree that AI will never make a human physician expendable (n=3,133; Bisdas et al., 2021) • AI Enthusiast (Beneficial) o Interviews with study volunteers in the Netherlands found that they generally expected AI to perform accurately (n=14; Jeung & Huang, 2023). o Survey of medical and dental students in 63 countries found that 83.9% think AI will be revolutionizing for medicine and dentistry. This view was most strongly held by male students from a developed country (n=3,133; Bisdas et al., 2021). o Survey of emergency and trauma surgeons worldwide found that 86% thought AI will improve acute care surgery (n=200; De Simone et al., 2022). • Informed AI User (Empowered) o Users of mobile health apps in Switzerland reported that they were more likely to trust apps that included medical certification, anonymization of data, and were affiliated with a trusted hospital system (n=106; Baldauf et al., 2020). o Survey of medical and dental students in 63 countries found that 85.6% think AI training should be a core part of medical training curriculum (n=3,133; Bisdas et al., 2021). o Survey of the general US public observed that there are important differences in how social groups perceive AI. Sizeism, transphobia, ableism, sexism, racism, and other factors influence AI-related medical experiences (n=187; Lee & Rich, 2021). CONCLUSIONS • The proposed Four Persona Framework can interpret many cases of expressions of human attitudes in HCI studies, including examples related to medical education. • Human attitudes are often more complex than any one persona, and change over time. Therefore, further framework development to accommodate multiple personas with varying intensities would make this framework more robust (see quandrant graphic) . • A future direction for framework development would be a qualitative or mixed methods study to test the proposed personas, identify gaps, and their prevalence in specific medical student, professional, and community populations.