A new study published in the Journal of Applied Psychology suggests that generative artificial intelligence can boost creativity among employees in professional settings. But the research indicates that these tools increase innovative output only when workers use specific mental strategies to manage their own thought processes.
Generative artificial intelligence is a type of technology that can produce new content such as text, images, or computer code. Large language models like ChatGPT or Google’s Gemini use massive datasets to predict and generate human-like responses to various prompts. Organizations often implement these tools with the expectation that they will help employees come up with novel and useful ideas. Many leaders believe that providing access to advanced technology will automatically lead to a more innovative workforce.
However, recent surveys indicate that only a small portion of workers feel that these tools actually improve their creative work. The researchers conducted the new study to see if the technology truly helps and to identify which specific factors make it effective. They also wanted to see how these tools function in a real office environment where people manage multiple projects at once. Most previous studies on this topic took place in artificial settings using only one isolated task.
“When ChatGPT was released in November 2022, generative AI quickly became part of daily conversation. Many companies rushed to integrate generative AI tools into their workflows, often expecting that this would make employees more creative and, ultimately, give organizations a competitive advantage,” said study author Shuhua Sun, who holds the Peter W. and Paul A. Callais Professorship in Entrepreneurship at Tulane University’s A. B. Freeman School of Business.
“What struck us, though, was how little direct evidence existed to support those expectations, especially in real workplaces. Early proof-of-concept studies in labs and online settings began to appear, but their results were mixed. Even more surprisingly, there were almost no randomized field experiments examining how generative AI actually affects employee creativity on the job.”
“At the same time, consulting firms started releasing large-scale surveys on generative AI adoption. These reports showed that only a small percentage of employees felt that using generative AI made them more creative. Taken together with the mixed lab/online findings, this raised a simple but important question for us: If generative AI is supposed to enhance creativity, why does it seem to help only some employees and not others? What are those employees doing differently?”
“That question shaped the core of our project. So, instead of asking simply whether generative AI boosts creativity, we wanted to understand how it does so and for whom. Driven by these questions, we developed a theory and tested it using a randomized field experiment in a real organizational setting.”
The researchers worked with a technology consulting firm in China to conduct their field experiment. This company was an ideal setting because consulting work requires employees to find unique solutions for many different clients. The study included a total of 250 nonmanagerial employees from departments such as technology, sales, and administration. These participants had an average age of about 30 years and most held university degrees.
The researchers randomly split the workers into two groups. The first group received access to ChatGPT accounts and was shown how to use the tool for their daily tasks. The second group served as a control and did not receive access to the artificial intelligence software during the study. To make sure the experiment was fair, the company told the first group that the technology was meant to assist them rather than replace them.
The experiment lasted for about one week. During this time, the researchers tracked how often the treated group used their new accounts. At the end of the week, the researchers collected data from several sources to measure the impact of the tool. They used surveys to ask employees about their work experiences and their thinking habits.
They also asked the employees’ direct supervisors to rate their creative performance. These supervisors did not know which employees were using the artificial intelligence tool. Additionally, the researchers used two external evaluators to judge specific ideas produced by the employees. These evaluators looked at how novel and useful the ideas were without knowing who wrote them.
The researchers looked at cognitive job resources, which are the tools and mental space people need to handle complex work. This includes having enough information and the ability to switch between hard and easy tasks. They also measured metacognitive strategies. This term describes how people actively monitor and adjust their own thinking to reach a goal.
A person with high metacognitive strategies might plan out their steps before starting a task. They also tend to check their own progress and change their approach if they are not making enough headway. The study suggests that the artificial intelligence tool increased the cognitive resources available to employees. The tool helped them find information quickly and allowed them to manage their mental energy more effectively.
The results show that the employees who had access to the technology generally received higher creativity ratings from their supervisors. The external evaluators also gave higher scores for novelty to the ideas produced by this group. The evidence suggests that the tool was most effective when workers already used strong metacognitive strategies. These workers were able to use the technology to fill specific gaps in their knowledge.
For employees who did not use these thinking strategies, the tool did not significantly improve their creative output. These individuals appeared to be less effective at using the technology to gain new resources. The study indicates that the tool provides the raw material for creativity, but the worker must know how to direct the process. Specifically, workers who monitored their own mental state knew when to use the tool to take a break or switch tasks.
This ability to switch tasks is important because it prevents a person from getting stuck on a single way of thinking. When the technology handled routine parts of a job, it gave workers more mental space to focus on complex problem solving. The researchers found that the positive effect of the technology became significant once a worker’s use of thinking strategies reached a certain level. Below that threshold, the tool did not provide a clear benefit for creativity.
The cognitive approach to creativity suggests that coming up with new ideas is a mental process of searching through different areas of knowledge. People must find pieces of information and then combine them in ways that have not been tried before. This process can be very demanding because people have a limited amount of time and mental energy. Researchers call this the knowledge burden.
It takes a lot of effort to find, process, and understand new information from different fields. If a person spends all their energy just gathering facts, they might not have enough strength left to actually be creative. Artificial intelligence can help by taking over the task of searching for and summarizing information. This allows the human worker to focus on the high level task of combining those facts into something new.
Metacognition is essentially thinking about one’s own thinking. It involves a person being aware of what they know and what they do not know. When a worker uses metacognitive strategies, they act like a coach for their own brain. They ask themselves if their current plan is working or if they need to try a different path.
The study shows that this self-awareness is what allows a person to use artificial intelligence effectively. Instead of just accepting whatever the computer says, a strategic thinker uses the tool to test specific ideas. The statistical analysis revealed that the artificial intelligence tool provided workers with more room to think. This extra mental space came from having better access to knowledge and more chances to take mental breaks.
The researchers used a specific method called multilevel analysis to account for the way employees were organized within departments and teams. This helps ensure that the findings are not skewed by the influence of a single department or manager. The researchers also checked to see if other factors like past job performance or self-confidence played a role. Even when they accounted for these variables, the link between thinking strategies and the effective use of artificial intelligence remained strong.
The data showed that the positive impact of the tool on creativity was quite large for those who managed their thinking well. For those with low scores in that area, the tool had almost no impact on their creative performance. To test creativity specifically, the researchers asked participants to solve a real problem. They had to provide suggestions for protecting employee privacy in a digital office.
This task required at least 70 Chinese characters in response. It was designed to see if the participants could think of novel ways to prevent information leaks or excessive monitoring by leadership. The external raters then scored these responses based on how original and useful they were. This provided a more objective look at creativity than just asking a supervisor for their opinion.
“The main takeaway is that generative AI does not automatically make people more creative,” Sun told PsyPost. “Simply providing access to AI tools is not enough, and in many cases it yields little creative benefit. Our findings show that the creative value of AI depends on how people engage with it during the creative process. Individuals who actively monitor their own understanding, recognize what kind of help they need, and deliberately decide when and how to use AI are much more likely to benefit creatively.”
“In contrast, relying on AI in a more automatic or unreflective way tends to produce weaker creative outcomes. For the average person, the message is simple: AI helps creativity when it is used thoughtfully: Pausing to reflect on what you need, deciding when AI can be useful, and actively shaping its output iteratively are what distinguish creative gains from generic results.”
As with all research, there are some limitations to consider. The researchers relied on workers to report their own thinking strategies, which can sometimes be inaccurate. The study also took place in a single company within one specific country. People in different cultures might interact with artificial intelligence in different ways.
Future research could look at how long-term use of these tools affects human skills. There is a possibility that relying too much on technology could make people less independent over time. Researchers might also explore how team dynamics influence the way people use these tools. Some office environments might encourage better thinking habits than others.
It would also be helpful to see if the benefits of these tools continue to grow over several months or if they eventually level off. These questions will be important as technology continues to change the way we work. The findings suggest that simply buying new software is not enough to make a company more innovative. Organizations should also consider training their staff to be more aware of their own thinking processes.
Since the benefits of artificial intelligence depend on a worker’s thinking habits, generic software training might not be enough. Instead, programs might need to focus on how to analyze a task and how to monitor one’s own progress. These metacognitive skills are often overlooked in traditional professional development. The researchers note that these skills can be taught through short exercises. Some of these involve reflecting on past successes or practicing new ways to plan out a workday.
The study, “How and for Whom Using Generative AI Affects Creativity: A Field Experiment,” was authored by Shuhua Sun, Zhuyi Angelina Li, Maw-Der Foo, Jing Zhou, and Jackson G. Lu.
Leave a comment
You must be logged in to post a comment.