Imagine a room of one hundred people solving the same problem. Each opens the same AI assistant and asks for ideas. Every person leaves with a better answer than they would have written alone, yet the room leaves with nearly the same answer one hundred times. Nobody did anything wrong; each person made a sensible choice. The loss of answer variety only appears when you look at the whole room. I worry that AI is judged too much on what it does for one person and too little on what it does to a group's variety of ideas. Science Advances 2024 shows that AI‑generated ideas boost individual creativity but make stories more similar, reducing collective diversity. 800‑person experiment confirms the same trade‑off across many writers.
The timing of the request matters. A July 2026 preprint finds that asking AI for ideas cuts overall variety, while using AI to refine existing ideas keeps diversity. A 2025 preprint with 486 participants shows that when the AI rewrites ideas, quality rises but ownership and variety drop; when people lead and the AI reflects, both quality and variety hold. Non‑native English speakers added more variety, but AI prompting erased that advantage. The pattern is clear: AI can enhance personal output but narrows the pool of ideas for everyone.
Because ideas are a shared resource, each AI prompt consumes a slice of collective variety. The article proposes a balanced approach: use AI to refine rather than generate, and consider the group impact alongside personal gain. This perspective frames AI assistance as a tool that must be used thoughtfully to preserve creative diversity. **
Source: Towards Data Science · Summarized by HeadlinesBriefing