AI makes work 'optional'
A feature this week argues that emerging AI agents risk turning schoolwork into something students can outsource, which often shows up as passivity rather than classic cheating (theatlantic.com). The suggested classroom response is to shrink optionality: require visible products, pair screens with partners or hands-on steps, and close tasks with reflection that can’t be automated (theatlantic.com).
A school assignment used to be something a student had to push through alone. New artificial intelligence agents are turning more of that work into something a student can delegate with a few prompts, clicks, or copied files, and The Atlantic says the result often looks less like obvious cheating than quiet disengagement. (theatlantic.com) That shift is showing up in the numbers. RAND reported on March 17, 2026, that student use of artificial intelligence for homework rose from 48 percent to 62 percent between May and December 2025 among middle school, high school, and college students. (rand.org) Most students do not describe the common uses as cheating. RAND found that nearly 80 percent said using artificial intelligence to understand an assignment was not cheating, while 72 percent said the same about brainstorming and 67 percent said the same about looking up facts. (rand.org) That is why the classroom problem is getting harder to spot. If a student asks a bot for an outline, a draft, a slide deck, or a cleaner paragraph, the final document can still arrive on time and look polished even when the student did very little thinking. (theatlantic.com) The tools are also changing shape. Microsoft now sells Microsoft 365 Copilot “agents” that can automate common tasks or work on a user’s behalf, which is a business feature but also a preview of what students will expect software to do everywhere else. (microsoft.com) Schools are already reacting by adding friction back into the process. Stanford said in October 2025 that instructors were redesigning coursework to place “important constraints” on student uses of generative artificial intelligence so classes still build deep learning and critical thinking. (stanford.edu) The practical fix is not banning every tool on every screen. The practical fix is making more of the work visible while it is happening, so a teacher can see notes, drafts, intermediate steps, partner discussion, or a physical product instead of only a finished answer. (theatlantic.com) That changes the assignment from “turn in a result” to “show your path.” A worksheet can be copied by a bot, but a marked-up page, a lab setup, a live explanation to a classmate, or a short reflection on what confused you forces evidence that a human mind was actually present. (theatlantic.com) The warning sign, then, is not just plagiarism software catching a fake essay. It is a student who stops wrestling with a hard paragraph, a math problem, or a history source because software now makes that struggle feel optional. (theatlantic.com) RAND’s survey captured the same tension from the student side. In 2025, 67 percent of students said artificial intelligence harmed critical thinking, up from 54 percent earlier in the year, even as more of them kept using it for schoolwork. (rand.org) The next fight in schools is not over whether a chatbot can write a decent paragraph. It is over whether teachers can design classes where the valuable part of learning is the part a machine cannot quietly do for you. (theatlantic.com)
