ClinQuiry — Clinical Research, Taught in Sequence

Core summary

AI tools like ChatGPT and Claude can help brainstorm research questions, identify gaps, and refine ideas — but they can also fabricate references and generate plausible-sounding but wrong suggestions. This lesson teaches you how to use AI as a thinking partner, not a replacement for your judgment.

Detailed explanation

What AI can do well in early research planning: 1. Brainstorm multiple angles on a clinical problem: Give the AI a clinical scenario and ask for 10 possible research questions. It excels at generating diverse perspectives you might not have considered. 2. Help structure PICO: Paste a vague idea and ask the AI to format it into PICO/PECO with alternatives for each element. 3. Suggest gaps in existing research: Ask the AI what aspects of a topic have been understudied. Cross-check these claims with PubMed. 4. Refine wording: AI can help make your research question more precise and concise. 5. Generate counter-arguments: Ask the AI to critique your question using FINER criteria. What AI cannot do reliably: 1. Cite real papers: AI frequently invents references that do not exist. NEVER trust an AI-generated citation without verifying it on PubMed. 2. Confirm novelty: The AI does not have access to every published paper and cannot definitively tell you whether your question has been answered. 3. Guarantee accuracy: AI can generate confident-sounding but factually incorrect medical information. 4. Replace expert judgment: A supervisor, statistician, or domain expert must validate AI-generated ideas. The golden rule: AI is a brainstorming partner, not an oracle. Every AI suggestion must be independently verified before acting on it.

Clinical example

A resident types into ChatGPT: 'I work in a neonatal ICU. What are understudied research questions about kangaroo care?' The AI generates 8 ideas including 'effect of kangaroo care duration on gut microbiome in premature infants' — an interesting angle the resident had not considered. But the AI also cites a 2022 Lancet paper that does not exist. The idea is valid; the reference is not.

Research example

Sallam (2023) published a review in Cureus noting that AI chatbots can be useful for generating research hypotheses but warned that their tendency to hallucinate references poses a serious integrity risk if outputs are not verified.

Knowledge check

Q1. Which of the following is a SAFE use of AI in research planning?

Q2. What is 'AI hallucination' in the context of research?

Q3. If an AI chatbot cites a specific paper with authors, year, and journal, you can safely include it in your research proposal without checking PubMed.