The Sweet Spot: Why Moderate AI Use Unlocks Your Best Creative Ideas

There’s a Goldilocks principle hiding in your creative process, and it has nothing to do with porridge.

Recent research by Hsuan-Che Brad Huang at the University of British Columbia suggests that artificial intelligence can genuinely boost human creativity, but only if you use it just right. Too little engagement and you miss the fresh perspectives AI can offer. Too much and you risk undermining your own sense of competence and falling into apathetic reliance. The magic happens somewhere in the middle.

Finding the Goldilocks Zone

Huang’s hypothesis rested on a simple observation: human creativity is often constrained by our own experience and assumptions. We all develop mental grooves that direct our thinking in predictable directions. Teams tend to be more creative than individuals partly because they bring different perspectives to the table. AI, theoretically, could serve as an artificial team member offering novel angles without the human baggage.

But large language models have their limits. They’re essentially statistical tools that produce “average responses” to questions, absorbing biases from their training data while lacking the beautiful idiosyncrasies of human thought. They can also undermine our confidence, creating a kind of passive apathy that’s hard to shake.

So Huang designed experiments to test whether moderation was indeed key. In the first study, around 150 participants helped a fictional student named Mike brainstorm a business idea with just $10 of seed funding. Some used ChatGPT once, others four to six times, and a third group used it nine or more times during the 15-minute task. Expert judges rated the final ideas on novelty, utility, and business value.

The results were striking. Those in the middle group produced the most creative ideas. A second experiment with 319 participants confirmed the finding. Then Huang took the research to professional creatives, surveying fashion designers, visual artists, authors, and animators. Those reporting moderate AI use (around 4 or 5 on a 7-point scale) showed the strongest creative output according to their bosses’ assessments.

The Personal Cost of Outsourcing Thought

The theory works in practice too. When tasked with developing a movie concept in 30 minutes, using moderate AI prompts actually produced something coherent and intricate. The AI helped broaden the range of possibilities, while human judgment kept the narrative coherent and purposeful.

But here’s where things get uncomfortable. There’s something valuable lost when we outsource our thinking, even partially. The genuine struggle of creativity, the sensation of synapses straining to find new connections, the surprise when unexpected motifs emerge from your own experience and emotions, all of that gets compromised when you lean too heavily on AI suggestions.

There’s also a troubling finding from Grace Liu at Carnegie Mellon University: people quickly become reliant on AI and less persistent without it. While this may not be permanent, it raises questions about long-term intellectual development. If you want to build genuine expertise and keep learning, outsourcing the struggle might be counterproductive.

The Ethics of Enhancement

Beyond the personal development concerns lies the question of ownership. Passing off an AI collaboration as purely your own work raises ethical red flags. There’s also the plagiarism risk, where the algorithm’s suggestions might inadvertently echo others’ work in its training data.

The real issue might be philosophical. Genius, as the saying goes, involves more perspiration than inspiration. If you’re genuinely interested in mastering your craft, whether that’s writing, design, or problem-solving, the struggle itself is often where the real learning happens. That effort isn’t wasted time; it’s the entire point.

None of this is meant to judge those who create with AI. Many people can harness it without stunting their intellect. But if you want to keep growing, to genuinely own your creative output, and to experience that electric moment when your brain makes an unexpected connection all on its own, you might want to ask yourself: is outsourcing the hardest parts of thinking really what you want to optimize?

Source: New Scientist