Photo via Fast Company
A multi-institutional study led by researchers at Carnegie Mellon University, Oxford, MIT, and UCLA has uncovered a troubling short-term risk of artificial intelligence use. According to the research, individuals who depend heavily on AI assistants for just 10 minutes experience measurable declines in their ability to solve problems independently. The study tested participants on math and reading comprehension tasks, with half receiving AI assistance and half working solo. When the AI was suddenly removed, the assisted group's performance dropped by approximately 20% compared to the control group—a significant gap that emerged within minutes of losing their digital crutch.
The nature of AI interaction matters significantly, the researchers found. Participants who asked their AI assistant for direct answers experienced the steepest performance decline, with 61% of study participants admitting they sought complete solutions rather than guidance. However, those who used AI strategically—requesting only hints or clarifications—maintained performance levels comparable to those who never used AI at all. This distinction is crucial for Atlanta's expanding knowledge-work sector, where companies increasingly deploy AI tools across finance, consulting, and professional services.
These findings align with earlier research documenting cognitive risks from AI adoption. MIT researchers previously measured brain connectivity in essay writers and found that those using large language models showed significantly weaker neural performance over a four-month period. Similarly, studies in fields like medicine and knowledge work revealed that professionals who over-relied on AI became less capable of independent task completion. The pattern suggests a troubling cumulative effect: the more people depend on AI, the more their own capabilities atrophy.
For Atlanta business leaders integrating AI into workflows, the implications are clear. While AI tools offer legitimate efficiency gains, the research suggests organizations must establish guardrails around implementation. Rather than allowing employees to outsource thinking entirely, companies should train teams to use AI as a thinking partner—requesting guidance and verification rather than wholesale answers. The researchers caution that without intentional safeguards, widespread AI adoption risks eroding the very human judgment and persistence that distinguish excellent problem-solvers in competitive markets.
