Rohan Paul Shares Paper on AI Cognitive Dependence

Rohan Paul posted on X about a paper he described as from Harvard and many other top universities. The post states that a small increase in LLM adoption could trigger a much larger jump in cognitive dependence on AI. It treats dependence as a feedback problem where handing reading, writing, reasoning, or verification to LLMs weakens the habits that sustain those skills. The paper argues for protecting independent reasoning before a tipping point, after which simply reducing AI use may not restore prior levels of autonomy. The post includes an attached screenshot described in the source line as page 1 from the arXiv document.