Why You Ignore AI's Explanations — and What That Means for You

We've all clicked "OK" on a pop-up without reading it. That's exactly what happens with AI explanations, according to a new research paper. The study argues that AI systems designed to explain themselves are missing a key piece: psychology. People decide whether to read an explanation based on three questions: Will it help me act better? Will it make me feel better? Will I understand more? If the answer feels like "no," we skip it.

The problem is our brains are terrible at predicting those answers. The paper points to common blind spots like overconfidence, the illusion of control, and confirmation bias. If you already trust an AI's answer, you won't check its reasoning. If you distrust it, you'll demand more and more evidence, even when it just wastes your time. Either way, the explanation fails its purpose.

This becomes dangerous when AI takes its own actions, like driving a car, approving a loan, or managing your calendar. You may need to intervene, but only if you understand what the AI is planning. A system that just shows you a wall of text won't cut it. The authors want to shift design from making explanations available to making them sought — like writing a warning people actually want to read, not just tolerate.

For everyday users, this means AI makers need to meet us where we are, explaining in moments we care about, in ways that respect our mental shortcuts. It's a call for AI that understands your attention is precious — and doesn't waste it.