SHAPR: A Solo Human-Centred and AI-Assisted Practice Framework for Research Software Development
This paper introduces SHAPR, a conceptual practice framework that operationalizes Action Design Research principles to guide solo researchers in developing research software with generative AI while maintaining human accountability, reflection, and methodological rigour.
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 massive, complex castle in the middle of a dense forest. This castle isn't just for living in; it's your research project. You have to be the architect, the bricklayer, the interior designer, the safety inspector, and the historian all at once.
Now, imagine you've been given a magical assistant—a super-smart robot that can lay bricks, mix mortar, and even suggest blueprints in seconds. This is Generative AI. It makes building faster and easier, but it also introduces a tricky problem: Who is actually in charge? If the robot builds a wall, do you understand why it's there? If the wall falls, is it your fault or the robot's?
This paper introduces a new set of rules called SHAPR to help solo researchers navigate this exact situation.
Here is the breakdown of the paper in simple, everyday terms:
1. The Problem: The "Solo Builder" Dilemma
In the past, building research software (like a custom app or data tool) was often done by teams where everyone had a specific job. But in universities, most researchers work alone. They have to wear many hats.
- The Old Way: They would just start coding.
- The New Way: They use AI to write code for them.
- The Danger: Because the AI does so much work, the researcher might stop thinking deeply about why they are building things. They might just accept the robot's suggestions without understanding them. This is like letting a robot drive your car while you sleep; you might get to the destination, but you won't know the route, and if something goes wrong, you can't explain what happened.
2. The Solution: SHAPR (The "Solo Human-Centred AI Practice" Framework)
The authors created SHAPR (pronounced like "shaper"). Think of SHAPR not as a new tool, but as a mental checklist or a training manual for how to use that magical robot without losing your own brain.
It is built on top of an existing, well-respected method called Action Design Research (ADR). If ADR is the "Grand Theory" of building things to learn, SHAPR is the "Daily Instruction Manual" for doing it alone with a robot.
3. How SHAPR Works: The Three Magic Rules
Rule #1: The "Hat Switch" (Role Multiplexing)
Since you are alone, you have to switch between different "hats" constantly:
- The Architect: Decides what the castle should look like.
- The Builder: Puts the bricks together.
- The Inspector: Checks if the walls are straight.
- The Historian: Writes down why you made those choices.
The SHAPR Trick: Even though it's just you, you must mentally switch hats and say out loud (or write down), "Okay, I am now the Inspector, not the Builder." This stops you from just mindlessly accepting the AI's code. It forces you to pause and think, "Does this make sense?"
Rule #2: The "Robot is a Intern, Not a Boss" (Human-Centred Governance)
In SHAPR, the AI is treated like a very fast, very knowledgeable intern.
- The intern can draft a letter, suggest a design, or fetch a tool.
- But the Boss (You) must sign off on everything.
- If the intern makes a mistake, the Boss is responsible.
- The Metaphor: Imagine you are a chef. The AI is a sous-chef who can chop vegetables incredibly fast. But you are the one who decides the recipe, tastes the soup, and takes the credit (or the blame) if it burns. You never let the robot decide the menu.
Rule #3: The "Time-Travel Journal" (Traceability)
When you build with AI, it's easy to lose track of how you got from Point A to Point B. SHAPR requires you to keep a "Time-Travel Journal."
- You don't just save the final code.
- You save the prompts you typed to the AI.
- You save the reflections on why you accepted or rejected the AI's suggestion.
- The Metaphor: It's like keeping a video diary of your construction site. If someone asks, "Why did you build the tower that way?" you can play back the video and say, "I asked the robot for ideas, it suggested a spire, but I realized a dome was better for my research, so I changed it." This proves you were thinking, not just copying.
4. Why This Matters (The "So What?")
Without a system like SHAPR, solo researchers using AI risk:
- Losing their learning: If the AI does all the thinking, the student doesn't learn the skills.
- Losing credibility: If they can't explain how their software works, their research isn't trustworthy.
- Getting stuck: They might build something cool but useless because they didn't check the "big picture."
SHAPR ensures that you remain the master of your own research. It turns the AI from a "black box" that spits out answers into a "co-pilot" that helps you fly, while you keep your hands on the controls and your eyes on the horizon.
Summary Analogy
Think of Research Software Development as cooking a gourmet meal.
- The Solo Researcher is the Head Chef working alone in a kitchen.
- The AI is a high-tech food processor that can chop, slice, and mix instantly.
- SHAPR is the recipe book and the kitchen safety manual. It reminds the Chef: "Just because the machine chopped the onions doesn't mean you don't need to taste the sauce. You are the Chef. You decide the flavor. You write down why you added the salt. Don't let the machine take over the kitchen."
The paper argues that by following these simple rules, solo researchers can use powerful AI tools to build amazing things without losing their minds, their learning, or their scientific integrity.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.