The Middlesex University study of 1,923 people in Canada and the US found that six in ten said AI "did most of the thinking," and men were more likely to rely on chatbots. Passive acceptance of AI outputs correlated with lower confidence and a weaker sense of ownership of ideas. Participants who challenged or supplemented AI responses retained greater confidence. Researchers recommend active judgement and oversight when using large language models in workplace tasks.
Study: Relying on Chatbots at Work Can Weaken Confidence and Ownership of Ideas
If you feel like AI now does the thinking at your job, scientists say your confidence and even your sense of authorship could be slipping away without you noticing. Sebastian Gollnow/dpa
A new study published in the American Psychological Association journal Technology, Mind and Behavior suggests that heavy, uncritical reliance on AI chatbots in workplace tasks can undermine people's confidence in their own reasoning and weaken their sense of ownership over ideas.
Key Findings
Researchers at Middlesex University surveyed 1,923 volunteers across Canada and the United States and found that about six in ten participants reported that artificial intelligence "did most of the thinking" for them. Men were more likely than women to lean on chatbot outputs. The study reports correlations — not causal proof — linking passive acceptance of AI-generated content to reduced confidence in independent reasoning and a diminished sense of idea authorship.
If you feel like AI now does the thinking at your job, scientists say your confidence and even your sense of authorship could be slipping away without you noticing. Zacharie Scheurer/dpa
How the Study Was Conducted
Participants used a variety of commercially available large language models while completing simulated workplace tasks such as planning under uncertainty, multistep sequencing, and decision-making. The research encouraged people to use AI tools as they normally would, then measured self-reported confidence, perceived ownership of ideas, and trade-offs between task speed and depth of thought.
"The issue was not AI use itself but the degree of passive acceptance," said Sarah Baldeo of Middlesex University. She added that exercising oversight and "active judgement" tended to leave participants feeling more confident in their own reasoning.
Broader Concerns
The APA coverage also referenced a warning from University of Pennsylvania researchers about a potential phenomenon they call "cognitive surrender," in which routine acceptance of algorithmic answers may displace intuitive and deliberative thinking. "AI tools are not merely assisting decision-making; they are becoming decision-makers," the Penn team cautioned.
Practical Takeaway
The study suggests a simple behavioral guardrail: engage critically with AI outputs. Challenging or supplementing chatbot suggestions with personal research or alternative perspectives appears to preserve confidence and authorship. Organizations using AI in workflows should encourage active oversight, independent validation, and transparent attribution of ideas.
Help us improve.
