AI Literacy Across Disciplines: Clinical Science

Insights and best practices for teaching AI literacy to science researchers

Posted on September 21, 2026 in Blog Posts

At a glance

Discipline	Clinical and Translational Science Collaborative
Audience	Researchers
Instruction format	Consultation
Primary AI tools featured	Offline chatbots
Core AI literacy themes	Critical Thinking as Design, Data Integrity & Privacy, Ethical & Responsible AI Use, Bridging the Confidence Divide

Setting the stage

Case Western Reserve University’s (CWRU) School of Medicine (SOM) seeks to preserve critical clinical reasoning skills and align with the university’s broader Academic Integrity Policy, which considers unauthorized AI use a form of academic dishonesty and misrepresentation. The university also cautions that overreliance on AI tools may undermine active learning. Unless an assignment specifically requires AI, all submitted work must be the student’s own and not the product of generative AI. The NIH supports the development and use of AI in research, but stresses the need for responsible, transparent, and equitable practices. As part of the NIH-funded Clinical and Translational Science Collaborative (CTSC) at CWRU SOM, I support researchers in multidisciplinary collaboration and prioritize research integrity policies in all their work. While this article reflects on offline, customized chatbots AI in research and proposal development, the CTSC, the SOM, and CWRU champion opportunities to use AI ethically, furthering the academic community’s access to resources.

The use of AI at CWRU is sophisticated and advanced across the campus, including academic specialization in AI. However, many researchers are coming to AI without academic training or invested experience in this emerging field, creating a perceived access gap between those who consider themselves experts in AI and those who are distrustful of the interface and assume that AI is cheating. Many understandably believe they are providing their research to an AI “black box” that has no discernment of integrity. I seek to bridge this divide by promoting engagement with AI to enhance research while also promoting ethical use and data integrity. I also remind researchers that we have all benefitted from the development of information technology over our lifetimes.

As a proposal developer working with researchers and project teams, I support grant development exclusively in the Health and Human Services, particularly the NIH’s National Center for Advancing Translational Sciences. In addition to assessing proposals for alignment with NIH grant mechanisms, I offer a range of supports and can tailor consultations to provide practical, project-based development of offline chatbots that serve specific research needs. A database developer beginning in the 1990s, I learned programming by spending a year digitizing middle school libraries using Winnebago CIRC/CAT in the DOS-based Borland Turbo Pascal. Subsequent internships and paid opportunities allowed me to build other catalogs and learn additional programming languages. Throughout the 2000s, I delighted in these opportunities as more sophisticated technology connected resources across institutions. From inside those initial generation of black boxes with their glowing green prompts, a world developed that researchers still rely on. AI is another tool for rapid development of robust, user-friendly interfaces suitable for research, and I want to support reluctant researchers to engage with responsible and discrete opportunities to have AI work well for them.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• What 1 Million ChatGPT Conversations Reveal about Users
• GenAI in Information Literacy Sessions: Strategies for One-Shot Instruction
• Balancing User Needs and Technology after System Migration

🧑‍🏫 Sign up for Choice/Clarivate’s new, free micro-course on AI literacy. In “Beyond the Basics,” expect to learn more about connecting libraries to the AI ecosystem, integrating AI into pedagogy, and creating library workflow efficiencies.

---

Teaching AI Literacy

Anyone with basic computer skills and access to software tools can build a chatbot. Templates for chatbots generally include instructions for the chatbot, knowledge sources for the AI to use, and prompts for how to request information. You can name and describe the purpose of your bot and choose whether it searches the internet or not. Our institution offers them through their AI service portal and presets the platform with instructions that can be modified specifically for your purpose. Institutionally sanctioned chatbot templates for research can be discussed with university technology offices if a researcher has questions about interfaces.

An offline AI architecture can be built to reflect a personalized intellectual framework, and I encourage researchers to consider how a chatbot can serve as a personalized annotated bibliography or an organizing space for developing theoretical concepts. With a clear research outcome or question, chatbot instructions can include specific goals, not just generic queries. Embedded theoretical definitions with key concepts, definitions, and frameworks enables interpretation and information with specific research methodology.

For example, one project used MeSH (Medical Subject Headings) through the US National Library of Medicine. Headings can be downloaded and then added to a chatbot. MeSH term matching can help characterize research and support connections to other related fields. In addition, a chatbot can be written to recognize complex patterns (e.g., linguistic structures, discourse markers, theoretical constructs) rather than just surface terms. Bloom’s Taxonomy, for example, is an excellent tool to add to a bot when developing patient or focus group questionnaires for evaluation. A bot can also be programmed to summarize, annotate, and tag findings according to your research questions and theoretical lenses, building up an organized, thematically coded database as you work.

The most relevant part of building a chatbot is that researchers actively engage in the process, deploying already honed critical thinking into a new tool for supporting their work. Through the engagement in the development process, researchers can learn how AI can responsibly augment, but not replace, core research and reasoning skills. The process of typing directions for a chatbot to follow and then observing the results is an educational process that creates hands-on mastery of critical language selection and accurate prompt development. Finding tools online to support the process also educates a researcher on how many tools there are to choose from. Code languages become important, and understanding online environments like Github both informs researchers on how code is used and encourages them to engage in forums where they can find similar projects and communities.

Activities, Approaches, and Pedagogy

Building an AI bot is somewhat therapeutic for reluctant AI users. The process of building and configuring a chatbot to help translate research knowledge into design prompts changes an individual’s understanding of how AI works. Many people consider AI as an oracle where questions are asked and wisdom is given. Building a chatbot to serve a researcher’s intellectual interests creates a more generative and personalized relationship with AI possibilities.

That said, unless researchers are working in a system designed for Health Information Portability Act (HIPAA) compliance, a chatbot, even an offline one, is not appropriate for Protected Health Information (PHI). Any interface that includes PHI must be assessed by the institution to assure secure access and data safety. Therefore, no chatbot should be built with the intended inclusion of patient information.

Finally, it will take time to develop a reliable chatbot. The data that is used, the instructions that are written, and the prompts that are provided all need to be honed for accuracy and reproducibility. Like entering any new relationship, the discourse will have to develop a specific cadence that works to generate a predictable output. That is why starting with tasks and queries that a researcher already knows well can support the chatbot-building process.

Tailoring Instruction to Disciplinary Needs

There is no doubt that AI can become an unwieldy interface pulling from verifiable and unverifiable sources. Maintaining data integrity means supporting a researcher to build chatbots that pull from reliable data. For example, the Center for Medicaid Services (CMS) has valuable and publicly available datasets that can be used for analysis. However, users also need to be able to target the dataset with meaningful inquiries. A search for telehealth coverage in the CMS database will also need prompts around significant policy changes, like the Balanced Budget Act of 1997, which was the first significant federal legislation to open the door for Medicare telehealth reimbursement. Instruction on chatbot building must be aligned with the medical context, emphasizing clinical reasoning and critical inquiry needed for accurate and meaningful results.

Highlighted Tools and Technologies

I would suggest that most researchers have experience in a programming language, regardless to whether they recognize its role in their professional lives. Building chatbot instructions can be as simple as writing plain-language scripts for what you want it to do. However, with the use of more specific coding language, the project can become a more sophisticated interface. Python, for example, is an accessible language for data analysis. I generally use JSON because it tends to return more consistent results. In addition, I like to use Retrieval-Augmented Generation (RAG) instructions because those instructions teach your chatbot to verify information. Learning to structure bot instructions and configurations can test the results more specifically, ensuring transparency and reproducibility. I also discuss offline, project-specific bots as a practical alternative to cloud-based, general-purpose AI tools, in line with privacy and integrity concerns.

Key Questions for Learners

• What specific research problem or workflow am I designing this chatbot to support?
  
• What data sources will the chatbot use, and how will I ensure their credibility, relevance, and traceability?
  
• How do institutional policies, ethical standards, and regulatory requirements shape what this chatbot can and cannot do?
  
• How will I design prompts, instructions, or code so that the chatbot reflects my disciplinary reasoning rather than replacing it?
  
• How will I evaluate the chatbot’s outputs for bias, accuracy, and usefulness over time?

Reflecting on Impact

Successes include increased researcher engagement with AI as a research enhancement tool, greater awareness of ethical and policy considerations, and successful deployment of offline bots that improve research efficiency while maintaining compliance. Challenges involve overcoming initial resistance or misconceptions about AI, building resilience as the development process takes time and critical thinking, and providing them some navigational instincts for the rapidly evolving technologies and policies.

By encouraging researchers to find components of their work where AI can serve them goes a long way to instilling confidence in AI’s utility. Reminding researchers that we have all been working in interfaces that collect and organize online information for a generation also helps create a history and a future for its research capabilities. That said, AI requires clear guidance and ethical deployment that includes use boundaries that meet institutional expectations. It is valuable to explore the extent to which AI can be customized for research and appreciate structured approaches that embed their own disciplinary frameworks into bot design.

Advice for Other Librarians/Instructional Partners

• Start with institutional and disciplinary policies to set clear expectations
  
• Emphasize hands-on, project-based learning to demystify AI tools
  
• Encourage critical thinking about AI’s role, limitations, and ethical use
  
• Highlight the importance of transparency, documentation, and continuous evaluation
  
• Foster multidisciplinary collaboration to ensure that AI tools are both technically robust and ethically sound

Additional resources

Databases mentioned in the article:

• Centers for Medicare & Medicaid Services – Search Data
• NIH – Medical Subject Headings
• HIPAA Guidance Materials
• HHS Adherence to Policy and Regulatory Frameworks

Related Posts