Examining and Addressing Barriers to Diversity in LLM-Generated Ideas

Yuting Deng; Melanie Brucks; Olivier Toubia

cs.CYFeb 23, 2026

Examining and Addressing Barriers to Diversity in LLM-Generated Ideas

Authors: Yuting Deng, Melanie Brucks, Olivier Toubia

Organizations: Columbia Business School

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 (CoT) 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.

Figures & tables

Industry Context	Psychological Need	Product Form
$\bullet$ Strength & Muscle $\bullet$ Cardio & Endurance $\bullet$ Mobility & Flexibility $\bullet$ Body Composition $\bullet$ Recovery & Injury $\bullet$ Balance & Coordination $\bullet$ Mental Wellbeing $\bullet$ General Wellness $\bullet$ None/Utility Only	$\bullet$ Progress & Mastery $\bullet$ Personalization $\bullet$ Motivation & Discipline $\bullet$ Fun & Engagement $\bullet$ Social & Belonging $\bullet$ Convenience & Access $\bullet$ Self-Expression & Identity $\bullet$ Novelty & Curiosity $\bullet$ None/Functional Only	$\bullet$ Wearable $\bullet$ App / Software $\bullet$ Smart Equipment $\bullet$ Immersive Tech $\bullet$ Traditional Equipment $\bullet$ Apparel & Accessories $\bullet$ Consumables $\bullet$ Integrated Furniture $\bullet$ Subscription / Coaching $\bullet$ Outdoor / Environmental

Table 1

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Figure EC.1: Study 2 Fixation Analysis: Slope of Within-Individual Diversity Accumulation for Humans vs. Seeded LLMs)

Group	Total Categories	Unique Combinations	Total Ideas
Human (Study 1)	28	197	830
LLM Default (Study 1)	22	83	830
LLM Default + CoT (sequential)	23	105	830
LLM Persona (Study 3)	27	193	830
LLM Persona + CoT (sequential)	27	203	830

Appendix

Table EC.1: Appendix F Full Sample Comparison: Diversity Across LLM Conditions (Sequential Generation + CoT Revision)

Metric	LLM Default (Study 1)	LLM Default + CoT (Sequential)	$p$ -value
Total Unique Categories (10×10 Ideas)	$18.45\pm 0.11$	$19.40\pm 0.10$	$<.001$
Total Unique Combinations (10×10 Ideas)	$39.15\pm 0.32$	$41.92\pm 0.28$	$<.001$
Average Pairwise Distance (10×10 Ideas)	$2.04\pm 0.01$	$2.13\pm 0.00$	$<.001$

Appendix

Table EC.2: Appendix F Bootstrapped Diversity Comparison: LLM Default vs. LLM Default + CoT (Mean

\pm

SE)

Metric	LLM Persona (Study 3)	LLM Persona + CoT (Sequential)	$p$ -value
Total Unique Categories (10×10 Ideas)	$24.02\pm 0.13$	$25.04\pm 0.12$	$<.001$
Total Unique Combinations (10×10 Ideas)	$56.97\pm 0.51$	$57.64\pm 0.54$	$.370$
Average Pairwise Distance (10×10 Ideas)	$2.36\pm 0.01$	$2.35\pm 0.01$	$.177$

Appendix

Table EC.3: Appendix F Bootstrapped Diversity Comparison: LLM Persona vs. LLM Persona + CoT (Mean

\pm

SE)

Group	Total Categories	Unique Combinations	Total Ideas
Human	28	206	990
LLM Default (Temp = 1.0)	22	88	990
LLM (Temp = 1.5)	23	99	990

Appendix

Table EC.4: Temperature Sensitivity Analysis: Diversity Outcomes

Diverse Text Generation

Diverse text generation studies how to produce varied, useful language outputs without sacrificing coherence, accuracy, or task performance.

See all 50 papers

AI and Its Impact on Creativity and Diversity: An Empirical Study of LLM-Generated Product Ideas
Jul 30, 2026Christian Terwiesch, Lennart Meincke, Karan Girotra +3LLM Prompting Computational Creativity
Operations, Information and Decisions Department, The Wharton School, University of Pennsylvania · WHU – Otto Beisheim School of Management · Mack Institute for Innovation Management, The Wharton School, University of Pennsylvania +3
Training Reasoning Models to Generate Diverse Scientific Ideas
Oct 15, 2026Manya Wadhwa, Pranav Narayanan Venkit, Hiroaki Hayashi +3Diverse Text Generation Scientific Hypothesis Generation
New York University · Salesforce AI Research · Stony Brook University
Human diversity fuels collective creativity that large language models cannot simulate or sustain
Jul 29, 2026Mengchen Dong, Hiromu YakuraComputational Creativity Diverse Text Generation
Center for Humans and Machines, Max Planck Institute for Human Development, Berlin, Germany

Jul 30, 2026J/K move · Enter open · S save

AI and Its Impact on Creativity and Diversity: An Empirical Study of LLM-Generated Product Ideas

Christian Terwiesch, Lennart Meincke, Karan Girotra, Ethan Mollick+2

Operations, Information and Decisions Department, The Wharton School, University of Pennsylvania · WHU – Otto Beisheim School of Management · Mack Institute for Innovation Management, The Wharton School, University of Pennsylvania+3

LLM Prompting Computational Creativity

Open paper AskPDF arXiv

This research examines how well large language models, or LLMs, generate new product ideas for college students priced under $50. Across a series of studies, we identify key strengths and weaknesses of using LLMs for product innovation. Our first study shows that LLM-generated product ideas have higher average quality than human ideas, based on purchase intent, and are 7 times more likely to rank in the top 10%. Our second study shows that this AI-induced creativity boost is not explained by the LLM's more persuasive pitching skills. Our third and fourth studies identify a weakness of using LLMs for brainstorming: AI-generated ideas are less novel at the idea level and less diverse at the set level. In our fifth study, we analyze prior LLM-based creativity studies and find consistently lower idea diversity across all of them, demonstrating the generalizability of these findings. Our sixth and seventh studies investigate techniques to mitigate this diversity loss. We compare LLMs from different vendors and versions and find that more recent models generate more diverse ideas, though they still fall short of human-level diversity. We also demonstrate techniques that increase idea diversity almost to the level of human idea generation: pooling ideas across vendors; prompt engineering, including Chain-of-Thought prompting and injecting heterogeneous personas or constraints; and creative agents that broadly explore the solution landscape to restore diversity. Finally, in our eighth study, we show that exploiting the near-zero marginal cost of AI idea generation by scaling the number of ideas steadily improves coverage of the idea space, approaching human-level coverage. We conclude by presenting actionable recommendations for innovation managers who want to identify better new product ideas with the help of LLMs.

Abstract

Diverse Text Generation

Explore similar work

AI and Its Impact on Creativity and Diversity: An Empirical Study of LLM-Generated Product Ideas

Training Reasoning Models to Generate Diverse Scientific Ideas

Human diversity fuels collective creativity that large language models cannot simulate or sustain