Collaborative and AI-Supported Requirements Elicitation: An Empirical Study
This empirical study demonstrates that combining stakeholder collaboration with AI-supported synthesis yields higher-quality and more clearly perceived requirements artifacts than traditional methods or direct AI generation alone.
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 trying to build a custom house. Before a single brick is laid, you need to sit down with the architect, the builder, the future homeowners, and the city planners to figure out exactly what the house should look like. This meeting phase is called Requirements Elicitation. It's messy, full of different opinions, and often results in a confusing pile of notes that are hard to turn into a clear blueprint.
This paper is like a controlled experiment to see if a "super-smart AI assistant" can help make this messy meeting process smoother and produce better blueprints.
The Setup: Four Different Ways to Build the Blueprint
The researchers created a fake scenario: an organization that needs a new system to manage meeting rooms because their current email-and-spreadsheet method is a disaster. They asked four different groups to create a "requirements document" (the blueprint) for this new system.
Here are the four teams they tested:
- The "Old School" Team (Collaborative, No AI): Three people sat down, talked, argued, and negotiated the features of the room system. They then had to write the final blueprint by hand, organizing their own chaotic notes.
- The "Smart Platform" Team (Collaborative + AI): Three people sat down and talked, but they used a special digital platform called Strateegia. This platform guided their conversation like a game master. At the end, an AI tool (powered by GPT) listened to their chat and automatically wrote the final blueprint for them.
- The "Solo Robot" Team (AI Only): A researcher fed the problem description directly into an AI and asked it to write the blueprint immediately, with no human discussion.
- The "Transcript" Team (Discussion + AI): They took the raw notes from the "Old School" Team (Team 1) and fed them into the AI, asking the AI to turn those messy notes into a clean blueprint.
The Results: Who Built the Best Blueprint?
The researchers hired two expert judges to grade the blueprints based on how clear, complete, and consistent they were. Here is what they found:
- The "Old School" Team (No AI): Their blueprint was the lowest-rated. Even though the humans had great ideas, they struggled to organize their thoughts into a clean, professional document. It was like having a brilliant conversation but a messy notebook.
- The "Solo Robot" Team: The AI did a decent job on its own. It produced a better blueprint than the humans working alone, showing that AI is good at understanding a problem and making a list of features.
- The "Smart Platform" & "Transcript" Teams (The Winners): The highest-rated blueprints came from the teams where humans talked first, and then the AI organized the results.
- Whether the humans used the special platform or just wrote notes that were later fed to the AI, the result was the same: the best documents.
- The AI didn't just list features; it understood the context and nuance of what the humans discussed, turning a messy conversation into a structured, clear plan.
What Did the Humans Feel?
The people in the "Smart Platform" group (Team 2) said the process felt clearer and easier than the "Old School" group.
- Analogy: Imagine trying to build a puzzle in the dark versus having a light that shows you where the pieces go. The AI-supported platform acted like that light. It didn't do the thinking for them, but it helped them stay organized and know what to do next.
- Interestingly, the humans didn't feel like the AI made the work less mentally taxing; they still had to think hard. But the AI made the process of getting to the final document feel much less confusing.
The Big Takeaway
The paper concludes that AI shouldn't replace the humans in the room.
- The Human Role: Humans are the experts. They bring the context, the priorities, and the "why" behind the decisions. They are the ones who know what the meeting rooms are actually for.
- The AI Role: Think of AI as the ultimate secretary or editor. It is amazing at listening to a chaotic meeting, sorting through the noise, and writing a perfect, organized report.
In simple terms: If you ask an AI to guess what a house needs without talking to the owner, it might get the basics right. But if you let the owner and the builder talk first, and then ask the AI to write down what they agreed on, you get a perfect blueprint that captures everyone's needs without the confusion. The best results happen when humans provide the ideas and AI handles the organization.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.