SShortSingh.
Back to feed

67% of Students Say AI Harms Critical Thinking, Yet More Use It for Homework

0
·2 views

A December 2025 RAND Corporation survey of 1,214 students aged 12 to 29 found that 67 percent believe increased AI use damages critical thinking skills, up more than ten percentage points in ten months. Despite this concern, homework AI usage rose from 48 percent to 62 percent in just seven months, with middle schoolers jumping from 30 to 46 percent and high schoolers from 49 to 60 percent. Female students expressed greater concern than males, at 75 percent versus 59 percent. Analysts suggest students are not simply contradicting themselves but are responding rationally to a grade-driven system where opting out of AI puts them at a competitive disadvantage. A separate November 2025 survey by the American Association of Colleges and Universities found that 95 percent of faculty feared AI would increase student overreliance, and 90 percent believed it would weaken critical thinking.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

How Xerox's PARC Invented the Digital World But Rarely Profited From It

Xerox established the Palo Alto Research Center (PARC) on July 1, 1970, with a mandate to develop technologies entirely outside its core copier business. Over the following decade, PARC researchers produced landmark innovations including the GUI-based personal computer, the modern mouse, Ethernet, the laser printer, and the Smalltalk programming language. Despite this extraordinary output, Xerox struggled to commercialize most inventions — its GUI-based workstation, the Xerox Star, was priced at $100,000 and failed to find a mass market. A 1979 visit by Steve Jobs to PARC influenced the development of Apple's Macintosh, though Xerox had already attempted to bring similar technology to market. In 2023, Xerox donated PARC to research institute SRI International, closing a chapter on one of technology history's most prolific yet commercially unrewarded laboratories.

0
ProgrammingDEV Community ·

Five Markdown Techniques to Write Cleaner, More Readable GitHub READMEs

A developer who has reviewed hundreds of open-source GitHub READMEs has identified five Markdown techniques that improve readability and user engagement. These include using HTML collapsible sections to hide lengthy content, GitHub Flavored Markdown task lists to display project roadmaps, and reference-style links to keep paragraphs uncluttered. Mermaid diagrams embedded in fenced code blocks can visually represent system architecture without requiring separate image uploads. Properly aligned table columns and a live Markdown previewer that supports GFM rendering round out the recommended practices.

0
ProgrammingDEV Community ·

MERN Stack Learner Upgrades Todo App to Dynamic React Props on Day 115

A self-taught developer reached day 115 of their MERN stack learning journey by overhauling a personal Todo application. The key upgrade involved replacing hardcoded static data with a structured dataset array passed dynamically through React Props. The app now renders task entries from a parent component, making the interface fully data-driven rather than visually static. This shift reflects a foundational React concept — separating data from UI by passing structured objects through functional component layers.

0
ProgrammingDEV Community ·

74 API endpoints, 2,000 paid calls: 5 endpoints drove 75% of AI agent revenue

A developer running NetIntel, a pay-per-call API platform settled in USDC, analyzed roughly 2,000 real transactions across 74 endpoints over several months to understand how AI agents actually spend money. Just five endpoints accounted for 75% of all revenue, with a single text-to-structured-JSON transformation endpoint alone generating 40% of total earnings. Most of the remaining endpoints earned virtually nothing despite weeks of development effort, revealing that breadth of offerings added maintenance cost rather than value. The data showed that AI agents pay most for reliable data transformation — converting messy text into clean, typed JSON — rather than for raw data access, because guaranteed structure eliminates costly parsing steps and uncertainty. Based on these findings, the developer has shifted strategy to deepen and improve the few high-performing endpoints rather than continuing to build speculative new ones.