The Secret to Continuous Creativity With GenAI

Based on research by Jing Zhou (Rice Business), Yingyue Luna Luan (University of Queensland), and Yeun Joon Kim (University of Cambridge)

Key takeaways:

• Augmented learning happens when humans and GenAI learn to work together in ways that improve creativity over time.
• But repeated use alone is not enough. Human-AI creativity can plateau when users rely on GenAI mainly for quick responses or more ideas.
• “Idea co-development” — exchanging feedback, testing ideas and refining them together — can help sustain creative gains.
   

Once generative AI tools like Claude and ChatGPT hit the market, they changed the everyday mechanics of workplace creativity. Brainstorming that once took hours can now begin with a prompt and a machine-made flood of ideas.

But faster ideation is not the same as lasting creativity. Indeed, new research finds that collaborating with AI does not boost joint creativity over time, unless users change how they interact with it. 

Published in the journal Information Systems Research and co-authored by Jing Zhou, the Mary Gibbs Jones Professor of Management at Rice Business, the paper explores “augmented learning,” a process in which people and AI actively adjust to each other across creative tasks, gradually improving what they can produce together.

The paper shows that augmented learning does not happen simply because a human and AI are placed in the same workflow. To keep joint human-GenAI creativity from stagnating, the researchers found, users need to move beyond prompting for more ideas and learn how to develop better ones.

Why GenAI’s creative advantage fades

To find out if human-GenAI collaboration naturally matures over time, the researchers conducted three mixed-methods studies. In study 1, participants were asked to generate creative solutions for societal and environmental issues, such as climate change, over 10 consecutive rounds. Some participants worked entirely on their own, while others collaborated with a custom GenAI chatbot designed to simulate human-like dialogue. 

The results were revealing. Participants working alone became more creative through repetition, while AI-assisted participants plateaued. In each round, independent evaluators rated ideas for novelty and usefulness, and early on, AI-assisted participants scored significantly higher. But as the experiment continued across a series of tasks over time, their scores stayed statistically flat. By the seventh round, the human-only group had caught up.

“The creativity bottleneck has moved from generating ideas to developing them — the slower work of questioning, refining and reshaping ideas after they arrive.”

To understand why the advantage faded, the researchers analyzed hundreds of conversations between users and GenAI. Users were generally receptive to AI-generated ideas — that wasn’t the issue. The problem lay in how people used the tool: rather than co-developing better ideas with their AI, they simply chased more of them.

In the study, most participants used AI as little more than an idea machine — proposing a concept and waiting for a response, or simply prompting the tool for more. But that kind of transactional back-and-forth did little to improve joint creativity over time.

The inherent strengths of GenAI may help explain why. “Because these tools are already so good at producing varied ideas quickly,” Zhou says, “the creativity bottleneck has moved from generating ideas to developing them — the slower work of questioning, refining and reshaping ideas after they arrive.”

The harder work of idea co-development

That’s where the human role becomes more demanding. The researchers found that continuous improvement primarily happened when users engaged in something they call “idea co-development.” In this mode, humans and AI actively exchange critical feedback, combine concepts, reframe problems and refine a single idea into a more viable solution. Rather than ask AI to keep producing options, users engage in a set of strategies to make ideas better.

For example, a user might propose a concept, ask the AI to test its practicality, revise the idea based on that critique, impose constraints (e.g., budget, risks) to increase quality and efficiency, and then work with the AI to shape it into a stronger solution. 

The study revealed that when users gradually increased their engagement in this specific co-development strategy, their joint creativity improved significantly. 

Yet, the researchers found that without explicit instructions, engagement with this strategy declined over time, and creativity plateaued for humans using AI tools. So, in a final experiment, the researchers tested whether explicit guidance could facilitate augmented learning. A treatment group of participants received instruction in idea co-development, including how to exchange feedback and refine ideas interactively with AI, while a control group received no such guidance.

The intervention worked. The treatment group showed a significant improvement in joint creativity across tasks, unlike the participants who did not receive training in the strategy of idea co-development. By training users in how to treat AI as a creative partner, rather than a mere idea generator, Zhou et al. found participants produced stronger creative work over time.

How companies can train workers to create with AI

“Our research has major implications for companies eagerly integrating AI into their daily workflows,” says Zhou. “To get the most out of these tools, organizations need to assess how their employees are actually using them.” 

Leaders should train employees to move beyond simple brainstorming — teaching them the co-development strategies from Zhou’s research and encouraging them to bring their distinctly human strengths to the process: social judgment, ethical reasoning and emotional intelligence.

Companies deploying AI tools should design systems that push users toward deeper engagement — asking questions, prompting elaboration, rather than just delivering answers. And by training employees in the co-development strategies this research identifies, companies can turn GenAI from a novelty into a lasting driver of creative improvement.

Written by Scott Pett

Zhou et al (2026). “Augmented Learning for Joint Creativity in Human-GenAI Co-Creation,” Information Systems Research.