Examining and Addressing Barriers to Diversity in LLM-Generated Ideas
#LLM diversity #AI ideation #Chain-of-Thought prompting #Knowledge partitioning #Human-AI collaboration #Innovation ecosystem
📌 Key Takeaways
- LLMs generate less diverse ideas than humans due to fixation and knowledge aggregation mechanisms
- Chain-of-Thought prompting reduces fixation in LLMs by encouraging structured reasoning
- Ordinary personas improve knowledge partitioning by serving as diverse sampling cues
- Combining both approaches produces the highest idea diversity, outperforming humans
📖 Full Retelling
Researchers Yuting Deng, Melanie Brucks, and Olivier Toubia published a groundbreaking study on arXiv on February 23, 2026, identifying mechanisms that reduce diversity in ideas generated by large language models and developing interventions to address these issues, as concerns grow that widespread reliance on LLMs could homogenize ideation and undermine innovation at a societal level. The research draws on cognitive psychology to identify two key mechanisms that undermine LLM idea diversity: fixation at the individual level, where early outputs constrain subsequent ideation, and knowledge aggregation at the collective level, where LLMs create a unified distribution rather than exhibiting the knowledge partitioning inherent to human populations. Through four comprehensive studies, the authors demonstrated that targeted prompting interventions can address each mechanism independently, offering theoretically grounded solutions to preserve creativity in AI-generated content. The findings provide practical strategies for human-AI collaborations that can leverage AI's efficiency without compromising the diversity essential to a healthy innovation ecosystem, potentially transforming how organizations and researchers approach AI-assisted creative processes.
🏷️ Themes
Artificial Intelligence, Innovation, Human-Computer Interaction
📚 Related People & Topics
Innovation system
The concept of the innovation system stresses that the flow of technology and information among people, enterprises, and institutions is key to an innovative process. It contains the interactions between the actors needed in order to turn an idea into a process, product, or service on the market.
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
--> Computer Science > Computers and Society arXiv:2602.20408 [Submitted on 23 Feb 2026] Title: Examining and Addressing Barriers to Diversity in LLM-Generated Ideas Authors: Yuting Deng , Melanie Brucks , Olivier Toubia View a PDF of the paper titled Examining and Addressing Barriers to Diversity in LLM-Generated Ideas, by Yuting Deng and 2 other authors View PDF HTML Abstract: Ideas generated by independent samples of humans tend to be more diverse than ideas generated from independent LLM samples, raising concerns that widespread reliance on LLMs could homogenize ideation and undermine innovation at a societal level. Drawing on cognitive psychology, we identify (both theoretically and empirically) two mechanisms undermining LLM idea diversity. First, at the individual level, LLMs exhibit fixation just as humans do, where early outputs constrain subsequent ideation. Second, at the collective level, LLMs aggregate knowledge into a unified distribution rather than exhibiting the knowledge partitioning inherent to human populations, where each person occupies a distinct region of the knowledge space. Through four studies, we demonstrate that targeted prompting interventions can address each mechanism independently: Chain-of-Thought prompting reduces fixation by encouraging structured reasoning (only in LLMs, not humans), while ordinary personas (versus "creative entrepreneurs" such as Steve Jobs) improve knowledge partitioning by serving as diverse sampling cues, anchoring generation in distinct regions of the semantic space. Combining both approaches produces the highest idea diversity, outperforming humans. These findings offer a theoretically grounded framework for understanding LLM idea diversity and practical strategies for human-AI collaborations that leverage AI's efficiency without compromising the diversity essential to a healthy innovation ecosystem. Subjects: Computers and Society (cs.CY) ; Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC...
Read full article at source