…
Tech & Science
AI and Creativity: A Study Identifies an Optimal Zone for Moderate Use
Herz — Tech & Science Desk · · 30s summary · 2 min read
The essentials
A doctoral study conducted at the University of British Columbia in Canada identifies a "Goldilocks zone" for artificial intelligence use in creative tasks. Moderate usage produces the best ideas; too little or too intense AI use harms creativity. Two controlled experiments involving approximately 150 and 319 participants respectively confirm this finding. A survey of creative professionals—designers, authors, artists—yields the same conclusion: those reporting moderate AI use are rated as more creative by their managers.
- Moderate AI use (4 to 6 prompts) produces the best creative ideas
- Two controlled experiments (~150 and ~319 participants) confirm this result
- Creative professionals with moderate AI use are rated higher by managers
- Heavy AI use risks generating creative apathy, according to the researcher
- Large language models tend to produce statistically average responses
What we know
A doctoral study by Hsuan-Che Brad Huang, conducted at the University of British Columbia in Canada, identifies a "Goldilocks zone" for artificial intelligence use in creative tasks. Moderate usage produces the best ideas; too little or too intensive use harms creative quality. The results are reported by New Scientist.
In a first experiment involving approximately 150 participants, each had 15 minutes to propose a business idea for a fictional student with 10 dollars in capital. Participants who submitted between 4 and 6 requests to ChatGPT received the best ratings from expert judges on the novelty, usefulness, and commercial value of their idea—outperforming those who used only one prompt and those who submitted nine or more.
A second experiment conducted with 319 participants replicated and confirmed these same results.
~150
Participants in the first controlled experiment
319
Participants in the second replication experiment
4 to 6
ChatGPT requests in the most creative group
Results among creative professionals
A survey of creative professionals—fashion designers, visual artists, authors, animators, technologists, and influencers—yielded the same conclusion. Those reporting moderate AI use, scoring 4 to 5 on a scale of 1 to 7, were judged most creative by their managers.
Why does too much AI harm creativity?
According to Huang, large language models (LLMs)—artificial intelligence models with a very large number of parameters—are essentially statistical tools. They tend to produce an "average response," stripped of the idiosyncrasies characteristic of the human brain.
These models can also absorb biases from their training data. Overly intensive use also risks undermining the user's sense of competence and producing a form of creative apathy.
What remains uncertain
Available information comes from a New Scientist article; the modalities of scientific publication for this thesis are not specified in accessible sources. The generalizability of results to other types of creative tasks or AI tools beyond ChatGPT is not established by available data.
Topics
Related articles
Latest articles
Comments
No comments yet. Be the first to react.
