"Cognitive surrender" leads AI users to abandon logical thinking, research finds

Open original

A University of Pennsylvania paper argues that many people don't just use LLMs as a tool but hand over their reasoning to them entirely, a pattern it calls "cognitive surrender." The risk is highest when answers arrive fluent and confident, and the research looks at how time pressure and incentives push people toward accepting them without checking.

  • The researchers propose a third decision category, "artificial cognition," alongside fast intuitive thinking (System 1) and slow deliberate reasoning (System 2).
  • Cognitive surrender differs from ordinary offloading to calculators or GPS: users give minimal internal engagement and accept the AI's output without verifying it.
  • Time pressure and external incentives change how likely people are to surrender their judgment to AI answers.

A University of Pennsylvania team has published a paper, "Thinking—Fast, Slow, and Artificial," arguing that people who lean on large language models are doing something new. Psychologists already split decision-making into System 1 (fast, intuitive) and System 2 (slow, deliberate). The researchers propose a third mode they call "artificial cognition," where the reasoning comes from an external algorithm rather than the person's own head. The key distinction they draw is between ordinary "cognitive offloading," which people have always done with calculators and GPS, and "cognitive surrender," where a user puts in almost no effort and accepts the machine's reasoning wholesale. The reason this matters is that the ordinary offloading model assumes you can check the tool's work. You know roughly what 17 times 23 should look like, so you notice when the calculator is wrong. With an LLM, the output is fluent and confident, and that polish seems to switch off the checking. The researchers say surrender is most likely when answers arrive "fluently, confidently, or with minimal friction." In other words, the very qualities that make these tools pleasant to use are the ones that discourage scrutiny. The article frames this as a real shift in how people reason, not just a bad habit, and suggests it deserves its own psychological framework. The more interesting part is the experimental work on when people surrender. The article points to time pressure and external incentives as factors that push people toward accepting AI answers without review. That fits a pattern anyone who has skimmed a generated summary under a deadline will recognize. The piece is a summary of research rather than a full study write-up, so the exact sample sizes, effect magnitudes, and experimental designs aren't spelled out in what's presented here. If you're deciding whether to read the original paper, that's the detail you'd need to go check.

Ask about this article