A study by Hsuan-Che Brad Huang indicates that moderate engagement with AI tools can boost creative output, while excessive or minimal use may hinder the process.
Key facts
- •Hsuan-Che Brad Huang conducted research at the University of British Columbia on the relationship between AI usage and creative output.
- •Experiments found that moderate AI engagement leads to higher scores in novelty, utility, and business value compared to low or high usage.
- •A survey of professional creatives showed that those using AI at a moderate level (4 or 5 on a 7-point scale) were rated as having the most creative ideas.
- •A study by Grace Liu at Carnegie Mellon University indicates that users may become reliant on AI and less persistent without it.
- •AI models are noted for producing 'average' responses and potentially carrying biases from their training data.
Research conducted by Hsuan-Che Brad Huang at the University of British Columbia suggests that AI can serve as an effective tool for enhancing human creativity when used in moderation. Experiments involving business ideation and professional creative work found that individuals who engaged with AI at a medium level produced higher-quality results than those who used it very little or very frequently.
By the numbers
Experimental Findings on AI Engagement
In a study involving approximately 150 participants, researchers tested the impact of AI on creative problem-solving by asking subjects to develop business concepts with $10 in seed funding. Participants were divided into groups based on their level of AI interaction: those who used ChatGPT once, those who used it four to six times, and those who used it nine or more times. Expert judges evaluated the final proposals for novelty, utility, and business value, consistently finding that the middle group achieved the highest average scores. A follow-up experiment with 319 participants confirmed these results. Further research involved a survey of professional creatives, including fashion designers, authors, and animators. When comparing their self-reported AI usage on a scale of 1 to 7 against assessments provided by their employers, those who reported moderate usage—scoring 4 or 5—demonstrated the most creative output.
Potential Risks and Limitations
While AI can help broaden the range of ideas considered, researchers and observers note significant limitations. Large language models often provide average responses and may absorb biases from training data. Furthermore, reliance on AI can undermine a user's sense of competence and lead to apathy. A study by Grace Liu at Carnegie Mellon University, currently under peer review, suggests that individuals may quickly become reliant on AI, potentially reducing their persistence when working without the technology. Concerns remain regarding the ethical implications of AI collaboration, the risk of plagiarism, and the potential loss of personal cognitive development that occurs when individuals outsource their thinking.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by New Scientist.
