A recent study conducted by researchers from Microsoft and Carnegie Mellon University explores the impact of generative AI on critical thinking abilities in the workplace.

The study emphasizes that “when used improperly, technologies can lead to a decline in essential cognitive functions that should be maintained.”

As individuals increasingly depend on generative AI in professional settings, their focus tends to shift toward ensuring that the AI’s output is satisfactory, rather than engaging in higher-level critical thinking processes such as creating, evaluating, and analyzing information. The research suggests that when humans only step in to correct AI when its responses are inadequate, they miss out on valuable opportunities to develop their judgment skills and enhance their cognitive abilities, ultimately leaving them unprepared for rare exceptions when they need to think critically.

In simpler terms, excessive reliance on AI for decision-making could lead to a decline in our own problem-solving skills when AI encounters errors.

The study involved 319 participants, all of whom reported using generative AI at least weekly in their jobs. They were asked to provide three examples of how they utilized generative AI, categorizing their tasks into three main areas: creation (such as drafting standard emails), information (like researching a subject or condensing lengthy articles), and advice (seeking guidance or creating charts from existing data). Participants also evaluated whether they employed critical thinking in these tasks and whether AI assistance reduced or increased their mental effort. Additionally, they rated their confidence in their own skills, AI, and their capacity to assess AI-generated outputs.

Approximately 36% of participants stated they applied critical thinking to prevent negative consequences associated with AI use. For instance, one participant mentioned utilizing ChatGPT to compose a performance review but felt compelled to verify the AI’s suggestions to avoid potentially embarrassing situations. Another participant noted adjusting AI-created emails to align with their boss’s hierarchical cultural expectations. In numerous instances, respondents cross-checked AI-generated content with general online sources such as YouTube and Wikipedia, which might undermine the original intention behind using AI.

To effectively offset the limitations of generative AI, workers need to recognize the potential shortcomings of these technologies. However, not all participants demonstrated an understanding of these limitations.

The researchers noted, “Awareness of the possible adverse effects of GenAI responses can trigger critical thinking, but this only occurs if the user is mindful of such risks.”

Interestingly, the study revealed that participants who expressed confidence in AI tended to engage in less critical thinking than those who felt more assured in their own abilities.

While the researchers stop short of claiming that generative AI makes individuals less intelligent, the findings indicate that an over-reliance on such tools may diminish our independent problem-solving skills.