Picture two coworkers facing the same deadline for a client report. One opens a chatbot, types a quick prompt, copies the polished result, and pastes it into the document. The other writes a rough draft first, then asks the same chatbot to sharpen the language. Both submit strong work. But do they walk away feeling the same way about what they accomplished?
A study published in Scientific Reports suggests they probably do not. The research offers evidence that the way people lean on artificial intelligence, not simply whether they use it, shapes how confident, connected, and fulfilled they feel about their work.
The question behind the research
Generative AI tools can speed up writing, analysis, and other professional tasks. But the researchers were interested in a quieter set of consequences. When a machine takes over a task you used to do yourself, what happens to your belief in your own abilities, your sense that the output is really yours, and your feeling that the work matters?
Elena Hayoung Lee of the University of Southern California’s Marshall School of Business, working with Yidan Yin of Penn State’s Smeal College of Business and USC colleagues Nan Jia and Cheryl Wakslak, built their investigation around three psychological ideas drawn from decades of research on work.
The first is self-efficacy, meaning a person’s confidence that they can complete a task through their own effort. The second is psychological ownership, the feeling that “this is mine” about something you produced. The third is work meaningfulness, the sense that what you do is purposeful and valuable. The authors argue that all three tend to grow when people exercise control over their work and can see a clear link between their effort and the result. Heavy reliance on AI, they suggest, could weaken that link.
How the experiment worked
The team recruited working professionals from five occupations: consultants, data analysts, human resources professionals, managers, and marketers. After accounting for participants who dropped out or did not follow instructions, the main analysis included 269 people. Each completed two short, job-relevant writing assignments, such as drafting a press release or a delicate email.
For the first task, participants were randomly assigned to one of three approaches. One group worked with no AI at all. A second group, labeled “Copy and Paste AI,” used a chatbot to generate the content and submitted it without changes. A third group, labeled “First Human Then AI,” wrote their own draft first and then used the chatbot to review and edit it. The researchers treated this third approach as active collaboration, where the person stays in charge and the AI assists.
To keep the conditions honest, the team disabled the copy-and-paste function for everyone except the passive group, required the collaboration group to submit their human draft before the edited version, and asked participants afterward to report how they actually used the tool.
Then came a second step that gives the study some of its depth. After the first task, every participant completed a second writing assignment entirely on their own, with no AI allowed. This let the researchers see whether the effects of the first experience lingered once people returned to working independently.
What the analysis found
Right after the first task, the passive copy-and-paste group stood apart. They reported lower confidence in their ability to do similar work without AI, a weaker sense of ownership over the output, and lower meaningfulness compared to the no-AI group.
The collaboration group told a different story. Their scores on self-efficacy, ownership, and meaningfulness looked statistically similar to those of people who used no AI at all. In other words, drafting first and refining with AI appeared to preserve the psychological connection to the work, while passively accepting AI output appeared to erode it.
There was a twist on the enjoyment side. Immediately after finishing, the passive group actually reported the highest task enjoyment and the most satisfaction with their final product. Letting the machine do the heavy lifting felt good in the moment.
The lingering aftermath
The second, AI-free task revealed which effects stuck around. Psychological ownership bounced back: once people produced something with their own hands, their sense that the work belonged to them recovered. The authors suggest ownership is closely tied to the act of doing the work itself.
Self-efficacy and meaningfulness behaved differently. Participants who had relied passively on AI in the first task still reported lower confidence and lower meaningfulness even after completing the second task independently. The researchers interpret this as a sign that these deeper beliefs about one’s own capability and the value of one’s contribution may be slower to recover.
The early enjoyment advantage also reversed. After returning to manual work, the formerly passive group reported the lowest enjoyment and satisfaction of all three groups. The authors describe this as a possible contrast effect: once you have tasted the speed and ease of AI-generated output, doing the work yourself can feel more tedious by comparison. The collaboration and no-AI groups, by contrast, kept fairly steady enjoyment throughout.
Checking the pattern in real workplaces
Because a laboratory writing task is not the same as a full job, the team ran a follow-up survey of 270 working adults across a range of tasks beyond writing. Participants rated how much they relied passively on AI versus collaborated actively with it, and then answered the same psychological questions.
The patterns broadly lined up with the experiment. Passive reliance on AI was associated with lower self-efficacy, lower psychological ownership, and lower satisfaction. Active collaboration was associated with higher levels of those same measures. When participants imagined a day when AI was suddenly unavailable, the same relationships held. The authors note that this survey cannot establish cause and effect, since it measures associations rather than testing them through random assignment, but it suggests the experimental findings extend beyond a single writing task.
What it might mean for work
The researchers point out that avoiding AI altogether is becoming impractical, with many leaders now urging employees to use it as much as possible. They caution that this push, framed around efficiency, could nudge people toward exactly the passive habits their study links to weaker confidence and a thinner sense of meaning. For organizations, the authors suggest that training and tool design encouraging active collaboration, rather than passive automation, may help capture AI’s productivity gains while protecting workers’ sense of competence and connection.
A few caveats are worth keeping in mind. Participants used the AI tool in a separate browser window, so the team could not track the fine details of how people interacted with it. The “active collaboration” condition tested only one workflow, drafting first and editing second, and the authors acknowledge that other collaborative styles might produce different results. The study also did not measure participants’ baseline comfort with AI. Still, the work adds a specific texture to a broad debate, suggesting that the question worth asking may not be whether to use AI, but how.
