AI creativity depends on the mix
The paper cuts through the usual yes-or-no fight about AI creativity. A system can make one artifact more novel while making the crowd more alike. The design question is composition: which humans and models are present, in what share, and how they connect.
What the source reports
Youngblood and coauthors argue that generative AI’s culture effect depends on mixed groups. People and algorithms act together in those groups. The paper separates novelty from diversity. Novelty is about one artifact. Diversity is about the whole group. The abstract says AI-assisted ideation often raises novelty in one person’s output. It can also narrow diversity across the group. That result is not fixed. Humans and models search in different but useful ways. So mixed groups can beat groups made only of humans or only of models, both on performance and diversity. The practical lesson is not that AI helps or harms by itself. Group composition shapes collective creativity.
Original source
- Title
- Collective creativity in hybrid societies
- Author
- arXiv, Youngblood, Mudd, Anglada-Tort, Jones, Miu, Omigie & Schedel
- Publication
- arXiv
- Date
- Wednesday, September 2, 2026
Also covered by the other newsletter.