How Creative is AI? Exploring Human and Generative AI Creativity Through the Circle Exercise

A Simple Challenge: Try This Before Reading On

Before you dive into this article, grab a pen and a piece of paper. Draw 20 empty circles in a grid. Now, in just 3 minutes, transform as many circles as possible into different objects—using the circle as an integral part of your drawing. Ready? Go! If you need more time please take it.

If you have just tried it, you have engaged in a classic creativity exercise, often used to assess divergent thinking. Now, here’s the real question: if an AI did the same exercise, how would it compare? Would it be more creative? Less? Or would it, like many of us, fall into familiar patterns?

At the Convergence Design Labs (Purdue University), we explored this exact question in a study on Generative AI and human creativity, using the Circles Exercise as a benchmark. What we found raises fundamental questions about how AI (thinks ?), how humans create, and whether AI can truly expand creative boundaries—or if it gets stuck in the same patterns as we do.

Human Creativity vs. Generative AI: What We Investigated

We wanted to understand narrow creativity—the tendency for both humans and AI to explore only a small subset of the vast design space of possibilities. Using a large dataset of human responses and AI-generated results, we analyzed:

1. How humans exhibit narrow creativity in design exploration.
2. How Generative AI (GenAI) performs the same creative task.
3. Whether prompting techniques, such as Chain-of-Thought (CoT), can help AI push beyond its limitations.

What we found was surprising—both (most) humans and AI tend to generate a limited range of ideas, sticking to conventional categories (like faces, sports equipment, and everyday objects). AI, despite its vast computational power, doesn’t automatically think “outside the box” unless guided with prompts into those new design spaces.

Key Findings: Where Do AI and Humans Get Stuck?

Humans and AI both struggle with "narrow creativity."

In the circle exercise, humans overwhelmingly drew objects from a handful of familiar categories, such as faces, sun, and wheels.

AI-generated results showed a similar bias—unless explicitly prompted to think differently, AI defaulted to common objects, just as humans did.

AI generates more but doesn’t inherently push creative boundaries.

While AI produced a larger volume of ideas quickly, it mostly stayed within incremental variations of known concepts.

For example, when prompted to create new uses for circles, AI often returned variations of clocks, buttons, or planets—expanding the quantity but not necessarily the novelty of ideas.

AI prompting techniques help, but only to an extent.

Using Chain-of-Thought (CoT) prompting, where AI is guided step by step to expand on its reasoning, led to some improvement in creative diversity.

However, even with CoT, AI still struggled to generate truly unconventional ideas without strong human guidance. Humans have to be creative about the prompting as well.

AI is better at structured ideation than open-ended creativity.

When given constraints or specific guidance towards creating abstract designs, AI performed better.

However, without such framing, AI’s creativity tended to gravitate toward known, safe, and frequently occurring patterns.

What This Means for the Future of Creativity and AI

Can AI actually be creative? The findings suggest that AI, like humans, gets stuck in familiar thinking patterns. This means that AI alone is not a creative force on its own —it requires human direction to push boundaries.

Does AI make us more creative? Potentially, yes. AI extends? "narrow creativity" as we called it by generating rapid ideas and offering new combinations. But without human intuition and intervention, AI remains constrained by its learned patterns.

The Future: AI as a Creativity Partner, Not a Replacement Rather than viewing AI as a creative replacement, we should see it as a tool for augmentation. The real breakthrough lies in humans learning how to work with AI to expand creativity—using creatively structured prompts, refining AI-generated ideas, leveraging AI for rapid exploration, and more importantly developing new methods in AI to support human creativity.

Recent research from a Microsoft team has also suggested that AI tools might be convenient for workers, but there is a good amount of risk they will become too reliant in the future. If AI starts helping users extend their creativity, it is critical that humans restrain themselves, without over relying on such tools.?

Many others have started to explore human-AI interactions for creativity: including Ethan Mollick from 美国宾夕法尼亚大学 - 沃顿商学院 Business School, Karl Ulrich Professor of Management from Wharton of University of Pennsylvania, and 谷歌 CEO Sundar Pitchai who suggested in passing in an interview that understanding hallucination as a means to help creativity would be an interesting direction.

Our findings suggest that we are severely limited in using AI as a tool for creativity by the creativity of human(s) using AI. In some ways I think the bi-directionality and the design navigation needs research into the black-box of such generative models and creativity tests themselves need to be understood for what they afford and constrain. Currently our creativity may be dependent on our abilities to prompt AI to then guide AI to new creative spaces and also what AI has been exposed to in its training.?Perhaps we should not use AI as a tool at all to develop our own creativity deeply so AI becomes a crutch and then everyone is developing similar thoughts with AI.

Takeaways: How Can You Apply This in Your Work?

• If you’re an educator or designer: Try using AI as a brainstorming partner but guide it carefully to avoid getting repetitive results. My suggestion is to use your own brain first before using AI.
• If you’re in product development or marketing: Use AI for rapid ideation, but layer in human intuition to push past predictable patterns.
• If you’re curious about AI’s limitations: Experiment with prompting techniques and see how structured or unstructured guidance affects AI’s creative output.

Final Thought: The Creativity Puzzle—Human + AI Together

Creativity is not just about generating more ideas—it’s about generating ideas that break patterns and challenge conventions. AI can help, but it still needs human intuition to lead the way.

Next time you face a creative block, try using AI—but also ask yourself: are you pushing beyond the familiar, or just iterating on the known?

While AI can enhance ideation speed, it does not naturally challenge existing patterns or break creative fixations. Humans provide intuition and abstraction, while AI provides speed and volume.

Some very creative toy designers have mentioned to me that they even use random TV channel browsing as a means to inspire themselves. And after decades of developing and teaching the toy design class and humans judging them that human creativity is unique and hard. Once judges see a creative design they tend to agree.

May be a good question is not how creative can AI be? How will AI impact human creativity in the future? Will humans be able to make AI be more creative? Let’s continue the conversation—How do you see AI shaping the future of creativity? Drop your thoughts in the comments! ??

Acknowledgments

This work was conducted in collaboration with my co-authors: Runlin Duan, Shao-Kang Hsia, Yuzhao Chen, Yichen Hu , Ming Yin, and Karthik Ramani . Their contributions and insights were invaluable in shaping the findings presented in this article. https://arxiv.org/abs/2502.07292

Disclaimer

The views, opinions, and interpretations expressed in this article are solely those of the author and any individuals quoted. They do not reflect the official policies, positions, or endorsements of Purdue University or any affiliated institutions. Any mention of research, projects, or affiliations is for informational purposes only and does not imply institutional endorsement.