Original title: After the Assistant Leaves: Cognitive Washout Dynamics and the Reversibility of AI-Induced Skill Decay
ROHIT YAJEE · September 2026
When people use generative artificial intelligence, their performance goes up, but their effort drops, and their critical thinking thins out. Studies show that using AI for just ten minutes leaves people worse at doing tasks alone and less willing to persist without help. But researchers do not know what happens next. The authors focus on this missing post-withdrawal period, which they call cognitive washoutThe process where skills lost during AI assistance fade away or recover after the AI is removed.. To study how skill loss behaves when help stops, they build a mathematical model to track recovery over time, how completely skills return, and relearning effort. They map out four possible paths: skills snap back completely, stay partly damaged, stay hidden until a stress test, or grow stronger through fresh struggle. Finally, they design a twenty-two week study with three arms across three domains, along with a small pilot, to test which outcome happens.
Imagine you always use a calculator for arithmetic. After a while, you can't do simple sums in your head. Now the calculator is taken away. Do you get better at mental math over time, stay bad forever, or only realize you're bad when a timed test comes? This paper studies what happens after the calculator leaves, and it predicts different outcomes: you might bounce back, stay low, hide the problem until a hard test, or even end up better than before from the struggle. In this paper, the calculator is the AI assistant, and the arithmetic is any skill you use it for.
If AI-induced skill loss doesn't reverse on its own, then students, professionals, and anyone relying on AI tools could be permanently worse at core skills. This matters for education, licensing, and product design—decisions made today assume the damage is trivial or fixable, but nobody has actually checked.
The abstract does not report any empirical results; it only proposes a framework and a protocol. The pilot stage is intended to establish feasibility, but no data is presented. The four outcomes and six predictions are derived but not tested. The paper does not claim to have measured washout in any real population.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
After the Assistant Leaves: Cognitive Washout Dynamics and the Reversibility of AI-Induced Skill Decay — ROHIT YAJEE — EdArXiv — doi:10.35542/osf.io/64qz7_v1
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.