Across three randomized experiments, about 10 minutes of AI assistance improved performance while available but led to more skipped and failed tasks once removed; participants who offloaded more showed greater declines in persistence. The authors recommend tutoring-style prompts and no-AI practice so AI augments rather than supplants human reasoning.
What Happened
On October 26, 2026, a study presented at the 2026 Conference on Language Modeling in San Francisco revealed that just ten minutes of AI assistance can erode motivation and independent performance. Grace Liu, a doctoral student in machine learning at Carnegie Mellon University, explained that “the concern is about what cognitive scientists call ‘desirable difficulties’ — the productive struggle that builds skill over time,” in an email to Live Science. Liu and her team recruited 1,222 participants across three randomized controlled experiments, each lasting 13 to 15 minutes. The experiments included two sets of mathematical fraction problems and a series of reading comprehension questions, with participants split so there was no overlap between trials.
The first experiment assigned 354 participants to either an AI chatbot group or a no‑AI control. Both groups received 12 identical fraction problems; the AI group could obtain immediate answers or prompts, while the control group worked unaided. After the AI was removed without warning, all participants faced three final fractions. The AI group solved more fractions and gave up less while the chatbot was present, but their performance dropped sharply once assistance vanished, leading to more skips and failures. A second experiment replicated this design with 667 participants and added a pretest to filter out those unable to handle basic problems; the pattern of decline persisted. A third experiment, involving 201 participants, focused on reading comprehension. Both AI and control groups answered five questions before the chatbot was disabled, after which each received three additional items. The results mirrored the math findings, showing that reduced persistence was not limited to arithmetic but affected comprehension as well.
“The most significant finding is that people who offload more have higher declines in persistence (increased skip rate), which is an important skill that cuts across domains,” Liu said. The researchers noted that immediate AI answers can make longer, self‑driven methods feel inefficient, leading users to abandon effortful tasks. Co‑author Brian Christian, a computer scientist at the Center for Human‑Compatible AI at the University of California, Berkeley, warned, “We need to work toward a future in which human abilities are, as much as possible, augmented rather than supplanted,” in a statement. The study, authored by Liu, Christian, Dumbalska, Bakker and Dubey, appears in the Proceedings of the Conference on Language Modeling (COLM) and is available as arXiv:2604.04721 [cs.AI].
The findings suggest that routine reliance on AI for simple reasoning can impair the very persistence needed for long‑term skill acquisition, raising concerns for education and workplace training programs.
What This Means For You
If you are a student or professional who regularly turns to AI for quick answers, the research shows that after just a few minutes of assistance your independent performance can suffer. Because the study demonstrated that removing AI caused a measurable drop in problem‑solving and an increase in skipping, you should limit AI use to complex tasks where it provides genuine insight, and reserve simple calculations or reading for self‑attempt. When you do use AI, ask for step‑by‑step guidance rather than direct solutions; the authors argue that tutoring‑style prompts help preserve the “desirable difficulties” essential for learning. Companies rolling out AI assistants for routine tasks should consider the findings, as discussed in AI in HR: Use Cases, Laws & Best Practices. Finally, after each AI interaction, take a brief “no‑AI” practice period to reinforce persistence and retention.
What We Don’t Know Yet
- Long‑term tracking: The study does not say how many months or years are needed to observe lasting declines in skill mastery.
- Real‑world impact: No data is provided on how these laboratory effects translate to everyday classroom performance or job productivity.
- Optimal AI design: While the authors recommend tutoring‑style assistance, the exact features that preserve motivation without removing challenge remain unspecified.
- Individual differences: The research does not disclose whether age, prior ability, or learning style moderates the degree of persistence loss.
Why It Matters
The results highlight a growing tension between AI’s convenience and the cognitive habits that underpin deep learning. By showing that even brief AI interactions can diminish the “desirable difficulties” that build expertise, the study adds empirical weight to concerns that over‑reliance on intelligent assistants may produce a generation of users who expect instant answers and struggle with sustained problem‑solving. This has implications for curriculum design, workplace training, and the ethical development of AI tools that aim to augment rather than replace human reasoning. Moreover, the call for tutoring‑style AI aligns with broader discussions about human‑centered technology, suggesting that future systems should prioritize scaffolding over substitution.
Key Takeaway
- AI assistance can quickly reduce persistence, causing users to skip or fail tasks once the tool is removed.
- Design AI to provide step‑by‑step instructions rather than ready‑made answers, preserving the productive struggle that builds skill.
- Regular “no‑AI” practice sessions help maintain motivation and independent performance.
- Educators and managers should monitor AI usage patterns to ensure that tools augment, not supplant, human effort.
Frequently Asked Questions
Does using calculators have the same effect as AI on persistence?
The source notes that calculators also allow people to offload computations, which could cause similar declines in independent performance. However, AI is unique because it offloads reasoning across many domains, not just arithmetic.
What kind of AI assistance is recommended to avoid eroding motivation?
The study argues that AI models and chatbots should be designed to give directions and instructions toward a solution, rather than the solution itself—more like a tutor. This approach preserves the “desirable difficulties” essential for learning.
How long does it take for AI assistance to impact motivation?
The research indicates that just ten minutes of AI assistance can begin to erode motivation, as observed in the experiments where
Leave a Reply