The overlooked AI risk: How automation could erode human critical thinking

There is a particular kind of risk that doesn't announce itself with a bang. It arrives quietly, in small surrenders—a decision you stop making because the algorithm makes it better, a calculation you stop performing because the machine does it faster, a judgment you stop trusting because the AI has been right more often than you have. Researchers are beginning to study this scenario seriously, calling it "gradual disempowerment," and they worry it represents a threat to human capability that is fundamentally different from the catastrophic AI scenarios that dominate public conversation.

The concern centers on a simple mechanism: as artificial intelligence systems take on more cognitive work, humans may progressively cede the mental effort required to think critically about those tasks. It is not that the AI fails or turns hostile. It is that we atrophy. We stop asking questions. We stop checking the work. We stop building the mental models that allow us to understand what is happening and why. The erosion is gradual enough that it may not feel like loss at all—it feels like convenience, like progress, like having more time for other things.

Jay Chooi, who co-founded and leads Robocurve, a firm that monitors AI-driven robotic systems, has entered this conversation as someone embedded in the technology itself. His company tracks how these systems operate in the world, which positions him to observe not just what the machines do, but how human behavior shifts in response. The question he and others are grappling with is whether we are heading toward a world in which the average person has become dependent on AI for thinking in the same way we became dependent on calculators for arithmetic—except that arithmetic is a narrow skill, while critical thinking is the foundation of informed citizenship, sound judgment, and human autonomy.

The distinction matters because it separates this risk from the more familiar doomsday narratives. A rogue AI system or a superintelligence that pursues misaligned goals would be a sudden crisis. Gradual disempowerment is slower and harder to see. It is the kind of change that happens at the level of habit and expectation. A student stops working through a problem because an AI tutor provides the answer. A manager stops reviewing data because the dashboard is always accurate. A voter stops reading policy because an AI summarizer tells them what to think. None of these moments feels dangerous. Each one feels rational.

What makes the scenario worth taking seriously is that it is not speculative in the way some AI risks are. We have already seen versions of it. GPS navigation has demonstrably weakened spatial reasoning in many people. Search engines have changed how we remember information—we remember where to find it, not what it is. Social media algorithms have reshaped how we encounter ideas, often without our conscious awareness. These are not catastrophes, but they are real changes in human capability and agency. The question is whether AI will accelerate and deepen these patterns to a point where something essential is lost.

The challenge for society, then, is not to reject AI or to pretend it will not reshape how we work and think. It is to build deliberate safeguards—structures and practices that preserve human agency and cognitive engagement even as machines handle more of the load. This might mean designing AI systems that explain their reasoning rather than simply deliver answers. It might mean creating spaces where critical thinking is protected and valued, where people are encouraged to question and verify rather than accept. It might mean recognizing that convenience and capability are not the same thing, and that a society of people who can think critically is worth more than a society of people who can simply delegate.

The conversation is still in its early stages. There is no consensus on how serious the risk is, or what the right response should be. But the fact that researchers are naming it, that people like Chooi are paying attention to it, suggests that the concern is gaining traction. As AI becomes more capable and more embedded in daily life, the question of what we preserve in ourselves—what we refuse to outsource, what we insist on thinking through ourselves—may become one of the defining questions of the next decade.