When teams brainstorm with AI, each person's ideas may improve while the whole group's ideas quietly start to look alike.
How does collective idea diversity differ when groups use AI assistance for ideation?
This explores what happens to the overall range of ideas a group produces (not each person's individual output) when members brainstorm with AI help, and whether the way AI is used changes the result.
This explores what happens to the overall range of ideas a group produces when its members brainstorm with AI help. Each person's ideas can get better while the group's ideas get more alike. The corpus suggests that the deciding factor is whether the AI generates the ideas or helps refine ideas people already had. In a preregistered experiment, AI-generated ideas reduced collective diversity for every kind of writer, while AI that only refined existing ideas left diversity intact Does AI assistance homogenize or preserve creative diversity?. That study also found that non-native English speakers added more diversity to the pool than native speakers did. AI ideation erased that advantage, so the people with the most distinctive perspectives lost the most.
It's easy to miss this effect, because the gains are visible and the losses aren't. AI help reliably makes writers produce more ideas with more detail, especially less experienced writers. But a study can report those gains without ever measuring diversity, so a conclusion about narrowed diversity can rest on missing evidence Does AI assistance actually narrow the diversity of ideas?. The same blind spot shows up in a different form. A single LLM's ideas can be rated more novel than human experts' ideas Do language models generate more novel research ideas than experts?. But novelty is judged one idea at a time, and diversity is a property of the whole set. An idea can look novel to the person reading it while hundreds of other people get something very similar.
The reason this happens sits below any single tool. Analysis of 70+ models across 26,000 open-ended prompts found an "Artificial Hivemind": different models, trained and aligned in similar ways, independently produce strikingly similar answers Do different AI models actually produce diverse outputs?. So handing different group members different chatbots doesn't restore variety. A more philosophical line in the corpus calls this a proliferation of claims without a proliferation of points of view Does AI generate diverse claims or diverse perspectives?. It argues that this sameness is harder to notice than mass media ever was, because each output feels tailored to you Does AI homogenize culture the way mass media did?.
The same pattern appears when the group itself is made of AI agents. LLM groups reproduce a familiar human result: discussion helps average members more than top performers. But they get there through more conformity, earlier convergence, and less sharing of unique information than human groups Do language model groups mimic human group reasoning patterns?. Diversity on its own isn't the fix either. Cognitively diverse agent teams beat solo ideation only when members have real domain expertise. Without it, diversity creates friction instead of insight Does cognitive diversity alone improve multi-agent ideation quality?.
The practical takeaway is about timing. Writers tend to reach for AI most at the very start, during ideation, and come back to it when they get stuck How do writers use AI through different creative stages?. That is exactly the stage where it does the most to make a group's ideas converge. If a group wants to keep its range, the corpus points toward generating ideas independently first and bringing AI in afterward to sharpen them.
Sources 9 notes
In a preregistered experiment, AI-generated ideas reduced collective diversity for all writers, while AI that refined existing ideas kept diversity intact. Non-native English speakers contributed more diversity than native speakers, but only AI ideation erased this advantage.
The paper's ideation experiment shows AI help increases idea count and detail, particularly for less experienced writers, but provides no diversity measure to support its conclusion about narrowed diversity.
A statistically significant study of 100+ NLP researchers found LLM-generated ideas rated as more novel than human expert ideas (p<0.05), though slightly lower on feasibility. Expert knowledge constrains novelty, while LLMs explore wider conceptual combinations.
INFINITY-CHAT analyzed 70+ models across 26K open-ended queries and found an "Artificial Hivemind" effect: models independently generate strikingly similar or identical responses due to overlapping training data and alignment procedures, undermining the diversity benefits of model ensembles.
Large language models generate numerous well-formed claims by following probabilistic patterns in training data, not by exploring competing argumentative positions. This produces volume without perspectival diversity—a thousand AI articles often represent approximately one viewpoint.
Show all 9 sources
AI mass-generates similar flows disguised as personalized outputs, suppressing novelty more deeply than pre-stamped commodities because contextual customization makes homogeneity invisible to individual users. Evidence: independent LLMs converge on similar outputs despite nominal competition.
LLM groups reproduce the human assembly-bonus asymmetry where discussion helps average members more than top performers, but achieve this through greater conformity, earlier convergence, and less unique information surfacing than human groups.
Multi-agent teams substantially outperform solo ideation, but only when members possess genuine senior knowledge. Diverse teams without expertise underperform even a single competent agent, because cognitive stimulation without expertise triggers process losses instead of insight.
An 18-participant study found writers use LLMs most intensively for ideation (generating initial ideas), then illumination (organizing thoughts), then implementation (drafting). Writers return to ideation during blocks, and unexpected outputs trigger new creative directions.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Human diversity fuels collective creativity that large language models cannot simulate or sustain
- Has the Creativity of Large-Language Models peaked? —an analysis of inter- and intra-LLM variability —
- What and Whose Knowledge? Measuring Epistemic Diversity in Large Language Models
- The Homogenizing Effect of Large Language Models on Human Expression and Thought
- What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity
- From Process Loss to Assembly Bonus: Human-Grounded Diagnosis of Multi-Agent LLM Collaboration
- Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers
- The Ideation-Execution Gap: Execution Outcomes of LLM-Generated versus Human Research Ideas