A study by Microsoft, in partnership with Carnegie Mellon University, highlights the effects of generative AI on knowledge workers.
The rise of generative artificial intelligence – commonly known as GenAI across the Atlantic – has transformed the way knowledge workers carry out their daily tasks.
A recent study by Microsoft Research and Carnegie Mellon University, titled “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers,” analyzes how these tools influence the critical thinking of people whose work relies on analyzing, managing, and applying intellectual knowledge, rather than performing manual labor.
The results, sometimes counterintuitive, suggest that trust in these tools profoundly shapes workers’ cognitive effort.
Assistive tools that change critical thinking
The study is based on a survey of 319 knowledge workers who shared 936 concrete examples of using generative AI in their work. The goal was to better understand how these technologies influence the way professionals mobilize their critical thinking.
First observation: far from being simply passive assistants, GenAI tools are transforming the very nature of critical reasoning. Indeed, users aren’t completely abandoning their analytical minds, but they’re redirecting them differently. Where previously they had to construct their ideas from scratch, they now find themselves in the role of evaluators and integrators of AI-generated content.
“What we’re seeing is a shift in the role of critical thinking,” the study notes. Rather than producing original reasoning, users are spending more time verifying, filtering, and re-adapting AI responses.
When using generative AI tools, the effort invested in critical thinking shifts from information gathering to information verification, from problem solving to integrating AI responses, and from task execution to task supervision.
When trust in AI reduces cognitive effort
One key insight from Microsoft’s study concerns the impact of trust on cognitive effort. The more trust a user has in AI, the less actively they engage in critical thinking. In other words, the more a person believes that AI provides relevant answers, the less they question its results.
This phenomenon could have worrying implications. If professionals blindly rely on GenAI tools without exercising critical judgment, it could lead to passive adoption of erroneous or biased content.
On the other hand, the study also shows that users with strong confidence in their own intellectual abilities continue to exercise critical thinking. They are more likely to question AI responses and refine their own conclusions.
AI tools appear to reduce the perceived effort required for critical thinking tasks among knowledge workers, particularly when they have high confidence in the AI’s capabilities. However, workers who are confident in their own abilities tend to perceive greater effort for these tasks, particularly when it comes to evaluating and applying AI responses.
Opportunities and risks for the future of work
The results of this research raise several questions about the evolution of cognitive skills in a world where generative AI is becoming ubiquitous. On the one hand, it offers considerable productivity gains by automating certain intellectual tasks and reducing the mental load on workers. On the other, it could lead to overreliance on these technologies and a weakening of independent reasoning abilities.
In this regard, our work suggests that generative AI tools should be designed to support knowledge workers’ critical thinking by taking into account barriers related to their awareness, motivation, and skills.
To avoid these abuses, researchers from Microsoft and Carnegie Mellon University emphasize the importance of designing tools that actively encourage critical thinking. Rather than fully automating certain tasks, AI should be designed to stimulate analysis and reflection. Features that encourage source verification or offer alternative perspectives could play a key role in this dynamic.
Towards harmonious cohabitation between humans and AI?
The study thus highlights a fundamental paradox: while AI is supposed to augment human capabilities, it sometimes risks weakening them if used indiscriminately. The key, therefore, lies in a balanced approach, where AI does not replace critical thinking but supports and enriches it.
As generative AI becomes an everyday tool, knowledge workers must cultivate a reflective and questioning approach to the content produced by these technologies. Trust is an asset, but it must never turn into blindness, the study notes.
With these lessons learned, companies and developers are therefore called upon to rethink the design of GenAI tools in order to make them true allies of human thought, rather than dangerous substitutes for intellectual effort, the researchers conclude.
