AI in Medicine Podcasts
- JAMA+ AI ConversationsThis collection of interviews with clinicians, researchers, and AI experts explores how AI is impacting medicine – from clinical practice to training and research.
- NEJM AI Grand Rounds PodcastNEJM AI Grand Rounds, a podcast from NEJM Group, explores the topics of technical, ethical, and legal questions around both the development and the sustainable deployment of AI while providing a forum to discuss evidence and resource sharing alongside debate about the best approaches to integrating AI within the healthcare system.
Guidance for Using Artificial Intelligence Resources
AI resources and tools should be seen as assistive tool in finding information, but not the end point of information retrieval. There are many benefits to AI along with potential drawbacks; see more in the Generative AI guide.
As with all medical information used for education and clinical purposes, critical appraisal of the provided resource is key. The information retrieved should be read, evaluated, verified with additional sources, appraised for context of your patient population, and understand its relevance in the growing body of acquired knowledge.
For more information on AI resources at UChicago:
CLEAR Framework for Prompts
Generative AI is particularly sensitive to prompts, so how a request is framed impacts the response. The CLEAR framework developed by Leo S. Lo provides a standard method for composing effective prompts.
CLEAR: Framework for Prompting
- Concise: Be specific and clear in your prompt. Adding too much information in your prompts can affect what a tool might focus on.
- Logical: Have structured and coherent prompt.s When writing a prompt, make sure that it follows a logical flow, especially when the prompt involves relationships between concepts or sequences.
- Explicit: Provide clear output specifications. Asking broad questions can lead to less precise answers. When creating a prompt, include relevant details that can help specify the type of response you want.
- Adaptive: Have flexibility and customization in prompts. When a prompt doesn’t get the response, you were looking for, be willing to adapt. A great thing about generative AI tools is that they often have a memory of previous prompts. You can use this to your advantage by asking follow-up questions or correcting and critiquing the tool’s response.
- Reflective: Adjust and improve your prompt by evaluating the previous responses. Taking the time after using a tool to reflect on responses can help you draft better prompts in the future and inform what types of inquiry work best with the tool.
Lo LS. The CLEAR path: a framework for enhancing information literacy through prompt engineering.” Journal of academic librarianship. 2023; 49(4).
Evaluating Responses
When using AI tools and resources, it is important to evaluate both the tool used and the response received.
Some guiding questions to use when evaluating a tool:
- What is the purpose of the tool?
- Who created the tool?
- How is it funded? Does that cause any bias?
- Is it free? If so, why?
- What was the tool trained on?
- What is the privacy policy?
- What are the Terms of Use?
Verify that the information and sources are accurate:
- Are sources cited? If so, are they real citations (not hallucinations) and accurate?
- Verify the content with other credible sources
- What is the goal of the response?
- In what perspective was it written from?
- How could bias have been introduced?
- If the response does not provide citations, consider using a different tool.
Additionally, to evaluate the response and the source citations provided by the generative AI tool, it is essential to critically appraise the information. The response and source citations should be considered in the context of your topic or patient population. Consider using an evaluation resource such as a critical appraisal checklist from CASP, BMJ, or JBI.
Subject Specialist
Joseph Regenstein Library, Room 263