SHAPR: Operationalising Human-AI Collaborative Research Through Structured Knowledge Generation
This paper operationalizes the SHAPR framework as a structured, traceable, and AI-executable system that integrates human-centred decision-making with AI-assisted development through iterative cycles and Structured Knowledge Units to enable rigorous, transparent, and scalable research practices.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a solo explorer trying to build a complex machine in a jungle. In the past, you had to chop wood, hammer nails, and test your invention entirely by yourself. Today, you have a super-smart robot assistant (Generative AI) that can chop wood, forge nails, and even suggest new designs for you in seconds.
But here's the problem: If you just chat with the robot and let it build things, you might end up with a working machine, but you won't know how it works, why you made certain choices, or how to rebuild it if it breaks. The process becomes a "black box"—magic that you can't explain.
This paper introduces SHAPR (Solo Human-centred and AI-Assisted Practice). Think of SHAPR not as a new way to build, but as a strict, organized diary and blueprint system for using your robot assistant. It ensures that even though the robot does the heavy lifting, you remain the captain, and every step of the journey is recorded so anyone can follow your trail.
Here is how SHAPR works, broken down into simple concepts:
1. The Two Workspaces: The "Chat Room" vs. The "Library"
SHAPR splits your work into two distinct places:
- The Chat Room (Interaction Workspace): This is where you talk to your AI. You brainstorm, ask for code, and try wild ideas. It's messy, fast, and conversational.
- The Library (Repository Workspace): This is your official record. It contains the final code, the official notes, and the "proof" of your work.
The Golden Rule: Nothing in the Chat Room counts as "research" until you move it to the Library. If you and the AI have a brilliant idea in the chat but don't write it down in the Library, it's as if it never happened. This keeps your work transparent and reproducible.
2. The Five-Step Loop: The "Recipe for Discovery"
Instead of just "coding and hoping," SHAPR forces you to follow a specific 5-step loop for every small improvement you make. Think of it like a chef tasting a soup repeatedly:
- Explore: "What if we add salt? What if we use a different pot?" (You and the AI brainstorm ideas).
- Build: "Okay, let's actually add the salt and change the pot." (You or the AI make the change).
- Use: "Let's taste the soup." (You test the new code or system).
- Evaluate: "Hmm, it's too salty. The texture is weird." (You analyze what happened).
- Learn: "Ah! I learned that too much salt ruins the texture. I will write this down." (You capture the lesson).
This loop turns simple coding into a knowledge-generating machine. You aren't just fixing bugs; you are learning why things work.
3. The "Knowledge Nuggets" (SKUs)
This is the paper's most creative invention. Every time you finish a loop (Explore → Learn), you extract a Structured Knowledge Unit (SKU).
Think of a SKU like a recipe card or a museum label.
- Context: What problem were we solving?
- Decision: What did we change?
- Observation: What happened?
- Insight: Why did it happen?
Instead of just saving the final code, you save these little "nuggets" of wisdom. Over time, you collect hundreds of these cards. Eventually, you can look at them and see patterns, like "Oh, every time I do X, the system slows down." These patterns become Design Principles—the rules that guide future research.
4. The Human is the Captain, The AI is the Crew
A common fear is that AI will replace researchers. SHAPR says: No.
- The AI is the crew member who can chop wood, carry heavy logs, and suggest routes. It is fast and powerful.
- The Human is the Captain. You decide the destination, you interpret the map, and you take responsibility for the final decision.
SHAPR ensures that even if the AI writes 90% of the code, the human must still explain why that code was chosen and what it means. This keeps the "human brain" in the loop, preserving the scientific integrity of the work.
5. The "AI-Readable" Feature
Here is the futuristic twist: SHAPR is written in a way that AI can read and follow it.
You can literally upload the SHAPR rules to your AI assistant and say, "From now on, you must help me follow these 5 steps and write these recipe cards."
- The AI becomes a "compliant assistant" that doesn't just guess, but follows a strict research methodology.
- It can even help you write the "recipe cards" (SKUs) automatically based on your testing results.
Why Does This Matter?
- For Solo Researchers: You don't need a big team to do big science. You can do it alone with an AI, as long as you keep your "diary" (SHAPR) organized.
- For Reproducibility: If someone else wants to check your work, they can look at your "Library" and your "Recipe Cards" to see exactly how you got your results. No more magic tricks.
- For Education: It teaches students that using AI isn't cheating; it's a tool. But the student must still understand the why and the how, not just the what.
In a Nutshell
SHAPR is a framework that turns the chaotic, fast-paced world of "AI coding" into a disciplined, scientific process. It treats every interaction with an AI as a potential experiment, records the lessons learned in "knowledge nuggets," and ensures that the human researcher remains the master of their own discovery. It's the difference between a random walk in the woods and a guided expedition with a detailed map.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.