A new study from Microsoft and Carnegie Mellon University suggests that over-reliance on generative AI may erode critical thinking skills. In other words: if you don’t use it, you lose it.
Researchers surveyed 319 knowledge workers across 936 AI-assisted tasks and found a troubling trend: the more users trusted AI-generated outputs, the less cognitive effort they applied. Said differently, confidence in AI correlates with diminished analytical engagement.
This is a textbook example of the automation paradox—offloading cognitive tasks to AI can make humans worse at them. Participants who blindly accepted AI’s suggestions reported weaker critical thinking skills, while skeptics remained more analytical, actively refining AI-generated content.
Another key finding: AI-assisted work tends to be more homogenous. Instead of injecting personal insight, many users defaulted to AI-generated solutions, especially under time pressure. Workers in high-stakes roles were more likely to scrutinize AI output, but those facing tight deadlines let AI take the wheel. This is one of my biggest concerns—AI-driven homogeneity could fast-track us to monoculture. ...
Shelly Palmer's 2/11/25 post.