Research Guides: Generative Artificial Intelligence (GenAI) in Higher Education: Write Better Prompts

Writing Better Prompts

CLEAR Framework: Concise, Logical, Explicit, Adaptive, Reflective
Complete sentences
Conversational: Talk to a friend, not a computer
Assign a persona:
- “You are a freshman college student...,“
- "You are a kindergarten teacher…,”
- “You are a forensic detective…”
And context: audience, goal, etc.:
- “You need to brainstorm keywords to research this topic…,”
- “You need to prepare a lesson on weather…,”
- “You need to generate questions for interviews based on the information gathered so far…”

Use descriptive, active words to clearly describe what you want but avoid extraneous words
Questions: specific but open-ended, avoid yes/no
Break task into logical steps when appropriate, with later responses building on earlier responses
- “First provide a synthesis of the information from these ten witness statements…”
- “Next generate approximately 2-5 follow-up interview questions for each witness based on the information known so far…”
Follow-up questions/prompts to clarify or refine
- “Questions 1 and 2 are good, but revise question 3 to not lead the witness towards…”

Plain-Text Version of Above:

Be mindful in your prompts...

Avoid confidential information such as financial details, proprietary information, or trade secrets.
Avoid sharing PII like full names, addresses, phone numbers, or any other data that could identify specific individuals.
Be cautious of sharing internal documents or details about company operations not meant for public disclosure.
If you use publicly available information, like a company website with a single employee, ensure you are not unintentionally sharing data that could easily identify the individual.
When using context from personal or private conversations, avoid quoting people directly. Instead, provide generalized, non-identifiable summaries.
If your input includes any legal terms or scenarios, ensure you are not violating any regulations or contracts, especially if bound by confidentiality agreements.
Avoid inputting sensitive or controversial topics in a way that could inadvertently misrepresent the context. Provide hypothetical scenarios instead.
Never enter passwords, access tokens, or any other login credentials into ChatGPT.
Eensure [sic]. you have permission to share any third-party data and anonymize it, when possible, to avoid breaching any data-sharing policies.
When inputting data for analysis, be mindful of any biases in the data. Ensure the questions or content you're inputting are neutral and factually accurate.