HEAR FROM THE EXPERTS
Suqing Wu and Yukun Liu
In our recent research, we set out to understand the trade-offs of working with generative AI in professional and learning contexts. Specifically, we explored what happens when people collaborate with generative AI on professional tasks and then return to working alone. We conducted four online experiments with over 3,500 participants from UK, asking them to complete activities such as writing emails, drafting performance reviews, and generating ideas.
The results revealed two sides of the story. Working with AI improved immediate performance: participants produced texts that were longer, clearer, and more creative compared to those who worked without assistance. However, these gains did not continue once participants switched back to solo tasks. After collaborating with AI, people reported feeling a stronger sense of control when working on their own, which suggested that their sense of control may have been compromised during the initial collaboration with AI; but they also felt less motivated and more bored. In other words, while AI can help people perform better at the moment, it may also make independent work feel less engaging afterward.
This pattern highlights an important tension. Generative AI can be a valuable partner for boosting short-term productivity, however it also has the potential to erode the psychological rewards that keep people interested and invested in their work. These findings matter not only for theory but also for practice.
For higher education and workplaces, the key lesson is balance. The challenge is not whether to use AI, but how to structure tasks so that its efficiency benefits do not come at the expense of the motivation and satisfaction that drive sustained learning and creativity. To the broader conversation, we believe preparing students for an AI-infused workforce requires not only training them to use AI tools effectively but also paying attention to how these tools shape and sometimes diminish their long-term intrinsic motivation.
While our findings were not unexpected, since we anticipated some motivational trade-offs once generative AI entered collaborative work, they have resonated strongly with both practitioners and scholars who recognized similar patterns in their own experiences. Other studies are beginning to report comparable concerns, suggesting this is not an isolated case. For example, a recent MIT Media Lab study found that individuals who frequently relied on ChatGPT showed reduced brain engagement and weaker linguistic performance over time, raising concerns about long-term cognitive impacts. Similarly, a new paper in Lancet Gastroenterology and Hepatology reported that clinicians, after prolonged exposure to AI-generated recommendations, became less motivated and less attentive when making independent decisions. Although contexts differ, these findings align with our core argument: AI can enhance efficiency but may also dampen engagement and motivation. Seeing our results echoed across such different fields has been both encouraging and thought-provoking for our team.
Readers can find our paper, “Human–generative AI collaboration enhances task performance but undermines humans’ intrinsic motivation”, in Scientific Reports. It is open access and freely available online through Nature’s website at: https://doi.org/10.1038/s41598-025-98385-2.
*********
Suqing Wu is an Assistant Professor at the School of Management, Zhejiang University, where her research focuses on organizational behavior, human–AI collaboration, and creativity in the workplace.
Yukun Liu is an Associate Professor at ShanghaiTech University, focusing on transformative work design, employee well-being, and human sustainability in the digital era.