How critical evaluation of artificial intelligence impacts work and study performance

Trust in AI and the role of critical thinking

A joint study by Microsoft Research and Carnegie Mellon University surveyed 319 professionals who use generative AI at least once a week and analyzed 936 real-world applications. The findings suggest that greater trust in the technology leads to less critical usage, while a more self-assured approach encourages deeper analysis.

Participants admitted they stop questioning AI responses when they lack the skills to review, improve, or guide the outputs. The study did not track changes in critical thinking over time, focusing only on the current situation.

When a tool becomes a crutch

An experiment involving nearly 1,000 high school students in Turkey, published in PNAS, compared math groups. During practical lessons, students using a ChatGPT-like interface performed 48% better than the control group, but scored 17% lower in assessments without AI.

Most of these students only requested ready-made answers and believed they had learned as much as their peers. Another group used a version that encouraged reasoning and achieved test results comparable to the control group. The authors argue that tools should be designed with safeguards to prevent passive use.

Practical application of critical thinking in daily life

According to the Microsoft study, critical thinking focuses on three key activities: verification, integration, and task supervision.

• Verification: Check numbers, names, and sources. Example: Confirm a sales figure cited in an AI-generated summary by referring to the original report.
• Integration: Adapt content to the context. Example: Rewrite a suggested email to match the tone and history of a specific customer.
• Task supervision: Define goals and criteria, and take responsibility for the final outcome. Example: Inform the AI about the target audience and desired report format, then review its response.

What remains unclear

The findings indicate a correlation, not a proven cause-and-effect relationship in the workplace. The Microsoft study relied on self-reported data, and the PNAS experiment involved teenagers in math classes.

Researchers suggest the results could help develop tools that promote critical thinking. In summary, actively verifying answers is linked to more critical technology use, while excessive trust without verification leads to tool dependency.