AI Tools Make Workers Less Critical, More Confident, Microsoft Study Finds

A study of over 300 knowledge workers has revealed a concerning paradox: the more confidence people have in AI tools like ChatGPT, the less likely they are to think critically about the AI’s output — even as they find their work easier to complete.

The research, published at CHI ’25 by researchers from Carnegie Mellon University and Microsoft Research, examined how professionals across diverse fields use generative AI tools in their daily work. Their findings suggest that while AI can boost productivity, it may also be eroding essential cognitive skills.

“Higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking,” the researchers found after analyzing 936 real-world examples of AI tool usage in professional settings.

This dynamic creates what the researchers describe as a “shift from material production to critical integration” – where workers spend less time creating content from scratch but need to invest more effort in verifying and refining AI-generated outputs.

For many professionals, AI tools have already become an integral part of their workflow. The study found that workers use AI assistants for everything from writing code and analyzing data to drafting emails and creating presentations. Nearly 97% of participants reported using ChatGPT, while other tools like Microsoft’s Copilot and Google’s Gemini were also popular.

The research identified three key shifts in how professionals engage with their work when using AI tools: information gathering becomes information verification, problem-solving transforms into response integration, and task execution evolves into task stewardship.

One participant, identified as P147, a lawyer, highlighted the verification challenges: “AI tends to make up information to agree with whatever points you are trying to make, so it takes valuable time to manually verify.” This experience was common across different professions, with workers reporting they need to carefully cross-reference AI outputs against reliable sources.

Time pressure emerged as a significant factor in how critically professionals engage with AI tools. A sales development representative in the study noted that meeting quotas often forced them to prioritize speed over verification: “The reason I use AI is because in sales, I must reach a certain quota daily or risk losing my job. I use AI to save time and don’t have much room to ponder over the result.”

The researchers warn that this pattern could lead to what they term “cognitive atrophy” – where regular opportunities to practice critical thinking skills diminish as workers increasingly rely on AI for routine tasks. This phenomenon mirrors earlier concerns about calculator use in mathematics education, but with potentially broader implications across professional domains.

The study also found that workers with higher self-confidence in their abilities tend to engage more critically with AI outputs, often using the tools to augment rather than replace their own judgment. These workers reported spending more time evaluating and refining AI-generated content, particularly in tasks requiring domain expertise.

Looking ahead, the researchers suggest that organizations need to develop strategies to maintain workers’ critical thinking skills in an AI-augmented workplace. They recommend designing AI tools that actively encourage critical engagement rather than passive acceptance of outputs.

The study’s implications extend beyond individual workplace productivity to raise broader questions about how AI tools might reshape professional expertise and decision-making capabilities in the long term. As these tools become more sophisticated and widely adopted, maintaining the balance between AI assistance and human critical thinking will likely become an increasingly important challenge for organizations and professionals alike.