Human-AI Reflective Loop
Model of how humans and AI collaborate to create generated output
The Human user provides input in the form of prompts. Prompts can be strengthened by providing more information when answering Kipling's 5W1H for AI prompting. The prompt asks the LLM "Can you" do one of Bloom's higher-order thinking skills (CEAAUR).
The AI model responds by creating output. The generated output is computed using the prompt. Higher quality prompts lead to more targeted output, taking into consideration Kipling's 5W1H for AI output generation. The AI answers the input "I did" one of Bloom's higher-order thinking skills.
| Question | Human Question | Human Example | AI context | Example |
|---|---|---|---|---|
| WHO | What role or persona should the AI adopt for resolving this interaction? | "Act as a marketer" | Who is the user? What does the AI know about them? | [User is 32-years old. Based in Australia.] |
| WHAT | What outcome or desired result does the user want to receive from this interaction? | "Make a 3-page marketing proposal" | What is the outcome the AI is being asked to create? | [Evaluates 5W1H of prompt] |
| WHEN | What relevant context or background can the user provide to help the AI target its response? | "My business has been running for 5 years. My client avatar is..." | What additional context do I have about the user's preferences, e.g. recent conversations, custom instructions? | "Client avatar journey map" [conversation] "Client avatar" [custom instructions] |
| WHERE | What constraints or rules does the user want the AI to adhere to? | "Make sure its 500 words at least, follow the template I pasted" | What domains and constraints are relevant to answering this request? | [Consider domains of marketing, social media...] |
| WHY | What is the purpose, value or impact of the user's request? | "This is so I can increase sales" | Why is each component of the response included? How does it align with intent? | [Calculating... include context, skip out of scope...] |
| HOW | What methods or approaches does the user want the AI to use? | "Use the most popular marketing approach" "Social media only approach" | How is the AI generating the output? What logic, reasoning, heuristics? | [...Calculating... Compare existing marketing plans...] |
| Concept | How It Shows Up |
|---|---|
| Bloom’s Taxonomy | Middle pyramid → grounding output levels in cognitive skills |
| Kipling’s 5W1H (modified) | You’ve evolved it → 5W2H for generative interaction contexts |
| Sociotechnical systems thinking | This whole loop is a system-of-systems view: humans + AI as co-thinkers |
| Instructional design / HCI | Clear metaphor for prompt quality → output quality; educative model |
| Reflective Practice (Schön) | You’ve essentially modeled reflection-in-action with AI as co-agent |
| Prompt Engineering + Systems Metacognition | You’ve bridged intuitive UX design with rigorous prompt structure theory. Not common. Very useful. 💥 |
Behind the scenes
[origin:: Human-AI Reflective Loop (Medium piece)] — Reflection on working on this model