Chain of Clarity Prompting: A New Approach to Structured Prompts
Although I’m generally a fan of unstructured, iterative prompting, sometimes a structured, detailed prompt is appropriate, for example when creating deep research reports. The quality of these reports is a direct function of the quality of the prompt used to generate the report. Because of the time it takes for AI to generate the reports (which can be very long, often over 30 pages), deep research doesn’t lend itself to iterative prompting. This makes it an ideal application for structured prompts.
Meta-prompting can be useful here, but I want to give you another tool, one that to the best of my knowledge is novel, Chain of Clarity (CoC) prompting. (Full disclosure: ChatGPT came up with this idea initially and we refined it together. I’ll write about the process in a later article.) CoC is a twist on chain of thought (CoT) prompting, which is a prompting technique that involves asking the AI model to explicitly show its reasoning, step-by-step, before giving the final answer.
Understanding Chain of Clarity
Chain of Clarity prompting is a structured, step-by-step approach to creating clear, focused prompts. The general idea behind CoC is to use combination of cognitive scaffolding and purposeful constraints to create effective instructions. There are three big “philosophies” that guide CoC:
Intentionality precedes insight: The quality of AI’s output depends on how clearly and purposefully you define your objectives.
Structure guides cognition: A good prompt structure guides AI’s “thinking” in much the same way a good outline or framework guides human thought.
Clarity is layered, not linear: Each part of the prompt, each layer, refines the focus through specific, intentional questions or instructions.
This will make more sense with a concrete example, but first I want to lay out the elements of CoC.
Goal: Clearly state your task, audience, and use case (purpose) of the output. The goal is the anchor and the first and strongest link in the chain. If this isn’t right, nothing else will be either.
Scope: List specific topics, theories, frameworks, etc. that you want AI to use. The scope is the link that tightens the focus; it defines the boundaries of the task.
Cognitive moves: Explicitly tell AI to perform critical tasks such as comparison, synthesis, or perspective-taking (using a particular perspective).The cognitive moves link is the engine. It describes how you want AI to think. Clarity becomes insight with the cognitive moves.
Format and style: Provide details about structure, length, voice and other formatting or stylistic characteristics. This is the usability link that ensures your output meets practical needs.
Meta and reflective checks: Ask AI to reveal assumptions, reflect on blind spots, and critically evaluate its own response. This forces AI to “think” more critically.
(Optional) Wildcards: Consider innovative twists such as changing perspectives, deconstructing concepts or forcing unrealistic constraints to gain novel insights.
Putting CoC Into Practice: A Deep Research Example
Now, let’s look at an example. We want to create a deep research report on how AI is affecting higher education.
Please create a Deep Research report following these parameters and instructions.
Goal:
Task: Write a deep research report on how generative AI is affecting higher education.
Audience: Faculty and university administrators
Use case: Inform institutional planning, curriculum design, and faculty training efforts.
Scope:
Topics: Teaching and learning, assessment of learning, risks, changing faculty roles, institutional trust
Framework/theories: Technological determinism, virtue ethics, stakeholder theory, constructivism, andragogy
Depth: Conceptually rich, but accessible to an educated audience; provide sufficient detail for a rich understanding
Cognitive moves
Compare and contrast: Juxtapose techno-optimist and techno-skeptic views; consider both utopian and dystopian views
Synthesize multiple sources: Integrate peer-reviewed research, industry reports, real-world cases, and faculty, student and administrative perspectives
Perspective-taking: Write from the viewpoint of a reflective, pragmatic educator navigating significant change
Format and style
Format: Briefing paper with clear sections, bullet points, and tables
Voice: Academic but conversational
Constraints: 10 - 15 pages
Wildcard options
Consider students’ points of view
When I gave ChatGPT Deep Research this prompt, it actually thanked me for the detailed prompt! You can view the entire chat session and report here.
I realize this is a detailed, complex prompt, at least on the surface. A closer look reveals that it’s really pretty straightforward once you understand the basic ideas of CoC prompting. The entire prompt looks intimidating, but when you break it down into the various sections, it isn’t nearly as scary. Just think of it as the sorts of detailed instructions you might give to a graduate assistant.
Refining Your Prompts Through Meta-Prompting
For critical tasks, consider combining CoC with meta-prompting by asking AI how the prompt can be improved:
Here's a deep research prompt. Can it be improved? Please create a Deep Research report following these parameters and instructions.
Goal: [rest of CoC prompt]
This works REALLY well. ChatGPT refined the prompt as follows:
Clarified Scope: The refined scope better focuses on specific areas (teaching, learning, assessment, faculty roles, risks, institutional trust).
Added Depth to Theoretical Frameworks: More specific alignment between theories and their relevance to AI integration helps direct research and analysis.
Perspective-Taking: Emphasizing the viewpoint of a reflective educator can lead to more practical recommendations for faculty and administrators.
Wildcard Options: The addition of "considering student perspectives" can prompt a more holistic investigation into how generative AI affects all stakeholders.
Formatting Details: Clearer direction on the paper’s structure, including visual aids, will help organize the content effectively for the audience.
The complete refined prompt is available in the chat session.
CoC isn’t suitable for every task. It works best for detailed, expansive tasks that don’t lend themselves to iterative refinement, especially if you need to do the same sorts of tasks frequently. As always, you should adapt the technique to what works for you. With a bit of experimentation, you can refine CoC to be a valuable technique in your AI prompting toolkit.
Want to continue this conversation? I'd love to hear your thoughts on how you're using AI to develop critical thinking skills in your courses. Drop me a line at Craig@AIGoesToCollege.com. Be sure to check out the AI Goes to College podcast, which I co-host with Dr. Robert E. Crossler. It's available at https://www.aigoestocollege.com/follow.
Looking for practical guidance on AI in higher education? I offer engaging workshops and talks—both remotely and in person—on using AI to enhance learning while preserving academic integrity. Email me to discuss bringing these insights to your institution, or feel free to share my contact information with your professional development team.
Nice work.