Presentation
Generative AI and Health Information: Applying the Technology Acceptance Model to Health Information in the Era of Generative AI
SessionPoster Session 1
DescriptionGenerative AI (GenAI) is rapidly transforming many sectors, including healthcare. This transformation goes beyond administrative support such as scheduling, billing, and record-keeping to include patient-facing applications, such as symptom checkers, triage assistants, and conversational health information tools. One especially promising application is addressing health literacy gaps.
Health literacy is a critical determinant of public health outcomes and disease prevention, yet limited health literacy remains a persistent challenge, contributing to preventable morbidity, mortality, and healthcare burden. Efforts to strengthen health literacy are therefore essential. One challenge lies in engaging young adults, who are less likely to proactively seek health information to make informed decisions that could prevent future diseases and illnesses. Physicians, meanwhile, often lack the time to provide adequate patient education, especially preventive guidance for individuals with fewer health concerns, like young adults. At the same time, this population increasingly turns to online platforms, social media, and emerging GenAI applications (e.g., OpenAI’s ChatGPT, Google’s Gemini) when they do seek health-related information. As a powerful conversational tool, GenAI offers promising access to accessible, personalized, and scalable health information, especially with domain-tuned models for healthcare. Given the potential of this technology to serve as a medium for health information dissemination, it is essential to understand how this demographic perceives and uses GenAI to design systems that effectively improve health literacy and promote better public health outcomes.
The Technology Acceptance Model (TAM) provides a theoretical framework that identifies perceived usefulness and ease of use as key drivers of technology adoption, while extensions of TAM emphasize trust as a critical factor in sensitive domains like healthcare. Large-scale initiatives such as Google’s Med-PaLM reflect the growing industry emphasis on building credible medical chatbots, yet the design choices that shape these systems, particularly their persona (e.g., doctor vs. peer) and communication style (e.g., formal vs. casual), remain underexplored.
Prior research with rule-based chatbots produced mixed results: some studies suggest that a sympathetic, peer-like persona and casual tone can build rapport and engagement, while others indicate that a more formal, expert-like persona and communication style foster credibility and confidence. These findings suggest that these design features may directly shape whether users trust the chatbot, perceive it as useful, find it easy to use, and ultimately choose to adopt it as a health information source. It is important to empirically examine the effects of such design choices in the more advanced context of GenAI-based chatbots.
The present study seeks to apply TAM by systematically examining the effects of persona (doctor vs. peer) and communication style (formal vs. casual) on young adults’ perceptions of trust, usefulness, ease of use, and adoption intentions in a GenAI health chatbot. 72 college students were recruited from the University of Central Florida. A 2 × 2 within-subjects design was employed. Each participant used predetermined questions to directly interact with four simulated GenAI chatbot conditions, presented in counterbalanced order. Chatbot persona was manipulated through names and avatars, while communication style was manipulated using OpenAI’s GPT-4o with one-shot prompts. The presented health information was held constant across conditions. Following each interaction, participants completed surveys measuring perceived usefulness, perceived ease of use, trust (including affection-based and cognition-based trust), and intention to use.
Results reveal that only chatbot communication style shaped user perceptions: a casual tone significantly improved affection-based trust, while a formal tone increased perceived usefulness and intention to use. Trust was a significant predictor of perceived usefulness, and usefulness, in turn, predicted intention to use in TAM. Additionally, qualitative responses provided explanations for participants’ intent to use each chatbot, highlighting tone and associated perceptions of credibility and perceived warmth among decisive factors.
Aligning with past literature, these findings demonstrate that design features, specifically communication style, can alter user perceptions and behaviors such as trust, usefulness, and adoption in GenAI applications for health information. The adaptability of GenAI means that future tools could dynamically shift between styles, tailoring communication to intended goals and contexts. Insights from this study can inform the development of effective and safe GenAI health tools for improving health literacy among young adults.
Health literacy is a critical determinant of public health outcomes and disease prevention, yet limited health literacy remains a persistent challenge, contributing to preventable morbidity, mortality, and healthcare burden. Efforts to strengthen health literacy are therefore essential. One challenge lies in engaging young adults, who are less likely to proactively seek health information to make informed decisions that could prevent future diseases and illnesses. Physicians, meanwhile, often lack the time to provide adequate patient education, especially preventive guidance for individuals with fewer health concerns, like young adults. At the same time, this population increasingly turns to online platforms, social media, and emerging GenAI applications (e.g., OpenAI’s ChatGPT, Google’s Gemini) when they do seek health-related information. As a powerful conversational tool, GenAI offers promising access to accessible, personalized, and scalable health information, especially with domain-tuned models for healthcare. Given the potential of this technology to serve as a medium for health information dissemination, it is essential to understand how this demographic perceives and uses GenAI to design systems that effectively improve health literacy and promote better public health outcomes.
The Technology Acceptance Model (TAM) provides a theoretical framework that identifies perceived usefulness and ease of use as key drivers of technology adoption, while extensions of TAM emphasize trust as a critical factor in sensitive domains like healthcare. Large-scale initiatives such as Google’s Med-PaLM reflect the growing industry emphasis on building credible medical chatbots, yet the design choices that shape these systems, particularly their persona (e.g., doctor vs. peer) and communication style (e.g., formal vs. casual), remain underexplored.
Prior research with rule-based chatbots produced mixed results: some studies suggest that a sympathetic, peer-like persona and casual tone can build rapport and engagement, while others indicate that a more formal, expert-like persona and communication style foster credibility and confidence. These findings suggest that these design features may directly shape whether users trust the chatbot, perceive it as useful, find it easy to use, and ultimately choose to adopt it as a health information source. It is important to empirically examine the effects of such design choices in the more advanced context of GenAI-based chatbots.
The present study seeks to apply TAM by systematically examining the effects of persona (doctor vs. peer) and communication style (formal vs. casual) on young adults’ perceptions of trust, usefulness, ease of use, and adoption intentions in a GenAI health chatbot. 72 college students were recruited from the University of Central Florida. A 2 × 2 within-subjects design was employed. Each participant used predetermined questions to directly interact with four simulated GenAI chatbot conditions, presented in counterbalanced order. Chatbot persona was manipulated through names and avatars, while communication style was manipulated using OpenAI’s GPT-4o with one-shot prompts. The presented health information was held constant across conditions. Following each interaction, participants completed surveys measuring perceived usefulness, perceived ease of use, trust (including affection-based and cognition-based trust), and intention to use.
Results reveal that only chatbot communication style shaped user perceptions: a casual tone significantly improved affection-based trust, while a formal tone increased perceived usefulness and intention to use. Trust was a significant predictor of perceived usefulness, and usefulness, in turn, predicted intention to use in TAM. Additionally, qualitative responses provided explanations for participants’ intent to use each chatbot, highlighting tone and associated perceptions of credibility and perceived warmth among decisive factors.
Aligning with past literature, these findings demonstrate that design features, specifically communication style, can alter user perceptions and behaviors such as trust, usefulness, and adoption in GenAI applications for health information. The adaptability of GenAI means that future tools could dynamically shift between styles, tailoring communication to intended goals and contexts. Insights from this study can inform the development of effective and safe GenAI health tools for improving health literacy among young adults.
Event Type
Poster Presentation
TimeMonday, March 234:45pm - 6:15pm EDT
LocationRhinelander Gallery
Digital Health