Prompting Frameworks

  • High-quality prompts eliminate ambiguity - Clear context upfront gets better responses
  • Use structured frameworks, e.g. the CLEAR framework, to ensure complete prompts
  • Never share sensitive data - No personal records, confidential research, or security information

High-quality prompts that eliminate ambiguity are crucial for getting useful responses from Large Language Models. By providing comprehensive context upfront, you’re more likely to receive output that matches your requirements.

Scenario: You have been asked to draft a recruitment call for participants in a bioscience study running at your department.

You write the initial prompt:

Write something about recruiting participants.

Problems with this prompt:

We can use some frameworks to standardize our queries and make sure they are complete and don’t forget any relevant details. There’s no right or wrong approach — different frameworks may work better for different AI tools and yourself, so use the one that works better for you.

The CLEAR Framework

The CLEAR framework offers a systematic approach to prompt construction that ensures comprehensive communication with AI models.

Context - Establish the situation and background

  • Explain who you are and what you’re working on
  • Provide relevant background information about your project or field
  • Include any important constraints or considerations

Length - Define output requirements

  • Specify how much content you want (word count, number of items, etc.)
  • Indicate format preferences (paragraphs, bullet points, sections)
  • Include any structural requirements

Examples - Illustrate desired outcomes

  • Show examples of the style or format you’re looking for
  • Provide sample inputs or scenarios when relevant
  • Include both positive examples (what you want) and negative examples (what to avoid)

Action - State the specific task

  • Clearly articulate what you want the AI to do
  • Use specific action verbs (create, analyze, summarize, compare)
  • Be direct and avoid ambiguous language
  • Focus on one primary task per prompt

Role - Define the AI’s perspective

  • Specify what expertise or viewpoint the AI should adopt
  • Indicate the audience the AI should write for
  • Guide the AI’s approach and decision-making framework

A good prompt example using the CLEAR framework:

Context: I am a postgraduate researcher in microbiology running a study on biofilm formation. We need to recruit 30 healthy adult volunteers to provide saliva samples under an approved ethics protocol. Participation involves a single 20-minute visit.
Length: Create a 250-350 word recruitment notice with a compelling headline and clear call-to-action.
Examples:
- Include a brief, friendly tone: “Help us understand how everyday bacteria form communities.”
- Avoid: overly technical jargon or anything that overstates benefits
Action: Write a participant recruitment notice to be posted on University noticeboards and student channels.
Role: Act as a researcher experienced in ethical participant recruitment who understands how to communicate clearly and avoid coercive or misleading language, while making the study sound interesting and worthwhile.

University of Bristol guidance

Independently of the prompting framework that you are using, never share these with AI tools:

🚫 Personal data: Participant records, staff information, health data
🚫 Confidential research: Unpublished findings, sequences or datasets, grant applications under review
🚫 Commercial sensitive: Partnership agreements, financial information, IP under patent
🚫 Legal privileged: Legal advice, disciplinary proceedings
🚫 Security sensitive: Passwords, system configurations, access credentials

Find more →



Exercise 3: Building Research Prompts

Scenario: You’re organizing a workshop on “Reproducible Data Analysis for Bioscientists” at the University of Bristol.

Task: Use the CLEAR framework to write a prompt in duck.ai that helps design the workshop agenda.

Requirements:

  • Use the CLEAR framework
  • Include appropriate constraints
  • Maintain academic integrity

Reflection Questions:

  • How do your prompts ensure AI enhances rather than replaces your expertise?
  • What verification steps would you take for each output?
  • How would you attribute AI assistance appropriately?

Don’t close the chat session after this exercise— you will use the same one for Exercise 4 in the next section.

Create a prompt to help design a workshop on “Reproducible Data Analysis for Bioscientists” at the University of Bristol.

Context: I’m a postgraduate researcher at the University of Bristol organizing a “Reproducible Data Analysis for Bioscientists” workshop for fellow PhD students from various bioscience disciplines with mixed coding backgrounds.
Length: Create a detailed 3-hour workshop agenda with 15-30 minute segments.
Examples: Include interactive elements like hands-on coding activities and small group discussions, similar to effective Carpentries-style workshops.
Action: Design an agenda that balances presentation, demonstration, and practice, progressing from basic to advanced concepts (version control, scripted analysis, environments, sharing).
Role: Act as an experienced workshop facilitator specializing in research software skills.
Additional Constraints:
- Maximum 20 participants to allow for individual attention.
- Must include time for networking and peer learning.
- Consider accessibility needs and different learning styles.
- Include evaluation/feedback collection methods.
- Build in buffer time for technical issues.