Tania Lombrozo, the Arthur W. Marks '19 Professor of Psychology at Princeton University, sees a dangerous cognitive shortcut that could soon grow worse. We routinely mistake other people’s knowledge for our own, and artificial intelligence systems may dramatically accelerate that illusion.
The trap starts simply enough. When Lombrozo’s daughter ran a fever, she consulted a pediatrician, who diagnosed the flu. She didn’t need to understand the underlying biology—how fever works or why hydration matters. The doctor’s explanation was enough: “fever because flu.” Problem solved. Her daughter recovered, and Lombrozo moved on without truly grasping the mechanism.
This division of labor sounds entirely reasonable. It is. Society has built extraordinary prosperity on this system, coordinating specialized knowledge so people can focus on what they do best. A person drives a car without knowing how to manufacture one. Communities make pins, cars, and newspapers through similar delegation. Medicine, science, engineering, and commerce all advance because we’ve learned to rely on expertise rather than master everything ourselves.
But the system carries a hidden cost. People consistently mistake specialists’ knowledge for their own genuine understanding. We feel competent because we know whom to ask, not because we comprehend the answer.
How Expertise Gets Confused with Understanding
Children as young as 3 to 5 years old already show this pattern. They know to ask a doctor about a broken arm and a mechanic about a flat tire. By ages 11 and 12, they’ve developed more sophisticated discrimination. In a 2004 study published in Child Development—now 22 years old—researchers asked why tennis balls bounce better on sidewalks than on grass. Most children recommended consulting someone who understood why bubble wrap protects glass, rather than someone who knew why tennis balls come in cans of three. They grasped the underlying physics principle, not just a surface similarity.
Evolution, culture, and personal experience shape whom people think they should ask. We develop genuine networks of expertise. The problem arises when we confuse knowing where to find an answer with understanding it ourselves. We’ve outsourced explanation without realizing it.
The AI Multiplication Problem
This illusion works acceptably when expertise remains scarce and people can verify the answers. You can check whether your mechanic fixed your car. You can see whether your child’s fever broke after the prescribed treatment. But artificial intelligence systems can produce plausible-sounding explanations at scale, and they may shatter this fragile equilibrium. When anyone can ask an AI system and get coherent answers on virtually any topic, the line between knowing something and knowing someone who knows something becomes dangerously blurred.
Lombrozo wrote “Why We Ask Why: The Science of Explanation and the Human Drive to Understand,” a work about why humans instinctively seek understanding beyond mere information. That drive—to comprehend rather than simply know where to find answers—may become harder to satisfy and easier to fake in an age of sophisticated AI systems.
Why This Matters:
The cognitive trap Lombrozo describes directly affects decision-making, institutional trust, and personal responsibility. As AI systems become more accessible and conversational, people will increasingly mistake algorithmic outputs for genuine comprehension. That creates vulnerabilities: people may make consequential decisions based on explanations they don’t truly understand, leaving them less likely to catch errors or question unsound premises. For markets, consumers and investors may operate on false confidence. For governance, citizens may become less able to hold institutions accountable because they confuse access to information with actual knowledge.
Modern prosperity rests on a division of labor, but that system depends on maintaining some connection between understanding and action. When the link breaks completely, accountability fractures with it. Policymakers should recognize that automation doesn’t solve every problem well, and that the illusion of understanding carries real institutional costs.