Skip to content
#generative ai #explainable ai Book Open access

Generative AI Makes Creative Output More Homogeneous

Sep 2026 · Proceedings of the 2026 European Conference on Cognitive Ergonomics · pp. 198-201 · 0 citations · 17 references

TL;DR

A systematic review and meta-analysis of empirical studies on homogenization in human–AI co-creation reveals a small but statistically significant homogenization effect associated with AI use.

Abstract

Does using generative AI make creative outputs more homogeneous? As creative practice increasingly involves using generative AI, concerns have emerged that creative outputs may become more homogeneous across users, even when individual creativity improves. We address this concern through a systematic review and meta-analysis of empirical studies on homogenization in human–AI co-creation (19 studies, 61 effect sizes). Results reveal a small but statistically significant homogenization effect associated with AI use (d =.334). The effect is robust across sensitivity analyses and not explained by publication bias. Rather than indicating a loss of creativity, the findings suggest a reorganization of creative diversity in human–AI co-creation. The results highlight that gains in individual creative performance may coexist with reduced diversity at the collective level.

Read PDF

Similar papers

#generative ai Book Open access Oct 2026

Combinatorial Capture: Generative AI and the Displacement of Transformational Creative Work

It is argued that current generative AI systematically favors combinatorial creativity while offering weaker support for transformational creativity, and that HCI should evaluate AI tools not only by the quality of creative outputs they enable but by the distribution of creative types they support, privilege, or suppre...

Peter Dalsgaard · 0 citations
#artificial intelligence Preprint Sep 2026

Collective creativity in hybrid societies

It is argued that creativity in the context of generative AI is best understood as a property of hybrid collectives, or populations of interacting people and algorithms, rather than of individuals.

Mason Youngblood, Katie Mudd, Manuel Anglada-Tort et al. · 0 citations
Case report Open access Sep 2026

Better Technology, Worse Motivation: Generative Artificial Intelligence’s Mediocrity Trap

This paper examines how generative artificial intelligence (AI) affects both productivity and motivation in creative work. In a randomized experiment, participants were asked to complete two illustration tasks with varying access to a text-to-image AI tool. By recording the production process and evaluating outputs eve...

Yvonne Jie Chen, Jie Gong, Jin Li et al. · 0 citations
Preprint Aug 2026

AI-AI co-creation outperforms human pairs in creative tasks

Prior research often finds that AI creativity is limited: single systems rarely outperform humans, and human-AI collaboration does not exceed human output. We argue these conclusions underestimate AI's potential because most studies do not allow iterative, multi-agent exchanges that mirror the social processes underpin...

Y. Luan, Luning Sun, YeunJoon Kim et al. · 0 citations
Open access Sep 2026

Creative potential of image-generative AI models for conceptual engineering design tasks

Abstract Creativity is paramount in design engineering, driving design innovation with new ideas. Visual representations are crucial in communicating and refining design ideas. The emergence of image-generative AI presents new opportunities, but a comprehensive understanding of AI creative capabilities, particularly in...

V. Chulvi, L. Ruiz-Pastor, A. Berni et al. · 0 citations

Related blog posts

Microsoft Research Blog Oct 7, 2026

Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses

Training AI agents with reinforcement learning can be challenging because their tools, context, and decision-making are managed by complex frameworks. Agent Lightning connects existing agents to RL training, making it easier to improve them without rebuilding them. The post Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.