Designing human and generative AI collaboration: Trade-offs in productivity, satisfaction, and creative diversity
PNAS Nexus · 30.10.2026 · 5 (11)
DOI: 10.1093/pnasnexus/pgag353 ↗Kurz gefasst
Fachartikel in der Zeitschrift PNAS Nexus, online erschienen am 10.11.2026. Verfasst von Kartik Hosanagar und 1 weiteren Autor:innen. Beteiligte Einrichtung: Operations, Information and Decisions Department, The Wharton School, University of Pennsylvania, Philadelphia, PA 19104, USA. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Rapid advances in generative AI have raised hopes for major productivity gains but recent research has also raised concerns about declining worker satisfaction and reduced creative diversity. While prior research has documented each of these effects individually, it remains unclear how collaboration design shapes the costs that accompany AI-enabled productivity gains. We address this question through a human-subjects experiment on a 1,000-word creative writing task, comparing three human–AI collaboration designs. Completion time was similar across the three designs but quality, satisfaction, and diversity differed across them. The Human Confirmation design, in which AI led ideation, outlining, drafting, and editing while humans primarily confirmed or rejected output, showed the clearest downsides: lower quality and satisfaction, and lower aggregate genre and semantic diversity. By contrast, designs that preserved human involvement in preproduction or throughout the writing process avoided the clearest quality and satisfaction losses without requiring longer overall completion times. Preserving active human involvement in preproduction was also associated with lower aggregate similarity in content than Human Confirmation. These findings show that collaboration design substantially reshapes trade-offs between efficiency, quality, satisfaction, and diversity, suggesting that the central challenge is not whether to adopt AI, but how to allocate creative control between humans and AI. That choice shapes not only the output of creative work but the experience of creative agency.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Operations, Information and Decisions Department, The Wharton School, University of Pennsylvania, Philadelphia, PA 19104, USA.
- Financial Planning, Housing and Consumer Economics Department, University of Georgia, Athens, GA 30602, USA.
ORCID-Profile: Kartik Hosanagar, Daehwan Ahn
Zitieren
APA
Hosanagar, K., & Ahn, D. (2026). Designing human and generative AI collaboration: Trade-offs in productivity, satisfaction, and creative diversity. PNAS Nexus, 5(11). https://doi.org/10.1093/pnasnexus/pgag353
Vancouver
Hosanagar K, Ahn D. Designing human and generative AI collaboration: Trade-offs in productivity, satisfaction, and creative diversity. PNAS Nexus. 2026;5(11). doi:10.1093/pnasnexus/pgag353
BibTeX
@article{Hosanagar2026,
author = {Hosanagar, Kartik and Ahn, Daehwan},
title = {Designing human and generative AI collaboration: Trade-offs in productivity, satisfaction, and creative diversity},
journal = {PNAS Nexus},
year = {2026},
volume = {5},
number = {11},
doi = {10.1093/pnasnexus/pgag353}
}
EUVIMED