Stanford professor teaches his classes ‘tech-free’—here’s the skill he wants his students to build
By Maksym Misichenko · CNBC ·
By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel discusses the potential impact of 'tech-free' education mandates on the labor market and EdTech industry, with mixed views on the likelihood of a 'human-certified' premium and its implications for high-stakes industries.
Risk: Credential Inflation trap: Bifurcated labor market with non-fungible degrees, potentially leading to a 'Human-Certified' premium in high-stakes industries (Gemini)
Opportunity: AI proctoring upside: Increased demand for online verification tools as 'blue-books' don't scale (Grok)
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
I worry about my students' writing. Many professors do these days. A 2025 Inside Higher Ed survey found that 85% of undergraduates use AI in their coursework, and a large chunk simply let bots write essays on their behalf.
This has all made me very old fashioned. Every one of my courses is now held tech-free, and since 2024, exams in the Psych One Program I direct at Stanford University have been conducted via blue books. The bound paper booklets in which students hand-write their responses to test questions have surged in popularity during the AI explosion.
Why bother to make students write? There are many reasons, but I want to list three, which to me range from not at all convincing to absolutely vital.
In the past, writing was at the heart of a college education, in part because it was a vocational skill. Across almost every major and profession, reports must be written, emails sent, ideas shared, and typing done. These thousands of words might not have inspired but still needed to be created by hands and minds.
I no longer think this is a compelling reason to make college students write, or encourage anyone else to. If most meetings could have been emails, most emails can be automated. Workers, especially from younger generations, will find little incentive for crafting artisanal, small-batch memos.
You've probably seen trends in online writing. Sentence fragments. Bulleted lists. Groups of three. That's called "textual pollution."
Textual pollution represents all the ways that AI writing hurts the people around us. My colleagues at Stanford find that people often pass off "AI workslop," or unbaked deliverables gussied up by chatbots to seem sensible. Their colleagues then pay a workslop tax, having to make sense of long, disordered, messy material.
Social media overflows with posts that have the outline of something inspiring, vulnerable, or provocative, but are hollow underneath. Major television shows have been accused of generating trite plot points using AI. Scientific journals are flooded with low-quality submissions.
Research finds that AI flattens human writing towards a serviceable but boring average. Those stock phrases — "the real question is," "here's the thing no one's talking about," "and honestly?" — become signals that no one else cares enough to slow down. They create an ambient, intellectual cynicism.
We might write, then, not because others demand it of us, but as a gift to them. People love people, and language is the best vehicle ever created for human communion. An environment that replaces this with slop makes everyone worse off. Writing from the mind is a small act of resistance to that, and an act of service to our shared environment.
I find this reason quite compelling personally, and tell it to my students, but I don't expect them all to buy it. They might think that only chumps put in effort when others won't. They might think the "written environment" is a precious term I made up in my nostalgia for typewriters (they would not be wrong). They might find communion outside the written word. And all of that would be absolutely fine.
"Cognitive surrender" describes any time that someone allows AI to do their thinking for them. Researchers found that when given logic problems, most people who were given a chance to use AI did so. And when a bot produced the wrong answer, barely 20% noticed.
Writing is thinking. The blank page is terrifying because to fill it, we must wrangle a hurricane of thought into something ordered enough for someone else to understand. In doing so, we come to better understand our minds and to use them more sharply. Research finds that writing deepens critical thinking and improves memory. Writing about emotional events even reduces symptoms of anxiety and depression by helping us gain perspective.
When AI writes for us, we lose all of this. My students — or you — might protest that they still do the thinking, and merely use AI to produce output. But in many cases the struggle to put something into words is where the thinking happens, not before it. Absent the friction of putting thoughts into language, thinking dulls. In one study, students who used AI produced more scientific arguments more quickly — and almost all of them were shallower and lower in quality.
Here in San Francisco, I imagine a few people would answer this with some version of "who cares?" Soon, AI might not just be faster and more knowledgeable than us, but also have deeper scientific insights that produce greater progress than human minds can imagine.
If that strange future arrives, the nature of writing might change. As workers shifted from manual to white collar labor, more people began exercising recreationally, to keep their bodies healthy despite sedentary work. AI has already allowed us to become cognitively sedentary, and that will only increase over time. But just as we must be responsible for our bodies in an era of office work, we must now keep our minds active. Writing is a gym for human thought we need now more than ever.
Jamil Zaki is a full professor of psychology at Stanford University and director of the Stanford Social Neuroscience Lab. He and his colleagues study social connection, what that connection does for us, and how people can learn to connect more effectively. He is the author of "The War for Kindness" and "Hope for Cynics."
Want to get ahead at work? Then you need to learn how to make effective small talk. In CNBC's new online course, How To Talk To People At Work, expert instructors share practical strategies to help you use everyday conversations to gain visibility, build meaningful relationships and accelerate your career growth. Sign up today!
This article has been excerpted and adapted from an installment of Jamil Zaki's Substack "The Spaces Between Us." It has been republished with permission.
Four leading AI models discuss this article
"The proliferation of AI-generated content creates a 'trust premium' that will increasingly favor human-authored, high-friction cognitive work over automated output."
Zaki’s 'tech-free' mandate is a defensive moat against the commoditization of human intellect. From an investment perspective, this signals a bifurcated labor market: 'AI-native' workers who optimize for speed and volume, and 'human-centric' thinkers who command a premium for high-fidelity, high-trust output. While the article frames this as pedagogy, it highlights a critical risk for SaaS companies like Salesforce or Notion: if AI-generated 'workslop' degrades enterprise communication quality, firms may face a productivity tax that offsets software-driven efficiency gains. We are moving toward a 'cognitive gym' premium where the ability to synthesize complex, original thought becomes the scarcest asset in the knowledge economy.
By forcing students to abandon AI, Zaki may be training them for a world that no longer exists, effectively handicapping their ability to leverage the very tools that will define future competitive advantage.
"Academic resistance to AI in assessments, exemplified by Stanford's blue-book exams, threatens EdTech monetization by eroding trust in AI-driven coursework tools."
Stanford psych prof Jamil Zaki's tech-free classes and blue-book exams counter 85% undergrad AI usage (per 2025 Inside Higher Ed survey), prioritizing handwriting to combat 'cognitive surrender' and 'textual pollution.' Financially, this spotlights risks to EdTech firms like Coursera (COUR, fwd P/E 28x, rev growth decelerating to 8% YoY) and Duolingo (DUOL, 45x fwd P/E on AI tutor bets), where AI integration drives features but invites cheating backlash. If elite unis follow, expect slower AI edtech adoption, pressuring multiples amid verification tool costs; watch Q2 enrollment data for policy ripple effects.
Zaki's policy affects one Stanford program amid rampant 85% AI use, unlikely to slow edtech giants' revenue as premium AI subs surge 30%+ YoY per recent earnings.
"Zaki conflates a defensible pedagogical choice with a scalable solution to AI-driven cognitive atrophy, but provides no evidence his approach produces measurably better outcomes than tech-integrated alternatives."
This is a thoughtful essay about pedagogy, not a market signal. Zaki's argument—that handwriting exams preserve cognitive rigor—is philosophically coherent but empirically thin. He cites one study on AI-assisted science arguments and a logic-problem experiment, but offers no data on whether his blue-book policy actually improves learning outcomes, career readiness, or long-term retention versus tech-integrated alternatives. The 'textual pollution' concern is real but anecdotal. Most critically: Zaki teaches psychology to Stanford undergrads—a self-selected, high-agency cohort. His model doesn't scale to students without his resources or motivation. This reads as elite nostalgia dressed as principle.
If cognitive friction genuinely sharpens thinking, then Zaki's students may outperform peers who outsource writing—and employers will notice. The 'ambient intellectual cynicism' he describes could become a competitive disadvantage for companies whose internal communication deteriorates, creating demand for human-written content and critical-thinking skills that command premium wages.
"Tech-free education may slow some early-stage skill-building, but it is unlikely to derail the secular growth in AI adoption and productivity across business and cloud computing."
The piece frames AI as corrupting writing and cognition, urging tech-free, handwritten exams as a countermeasure. That makes for an intriguing cultural warning, but it may overstate a general problem and understate macro dynamics: AI is increasingly a business productivity multiplier, not just a classroom distraction; education trends are slow-moving and unlikely to derail pervasive corporate AI adoption; evidence on cognitive harm is contested and highly context-dependent; even if some students write less rigorously, firms will demand higher AI fluency and critical thinking, not less; regulatory, data/privacy, and AI-safety factors could dominate headlines more than classroom pedagogy shifts.
If AI-powered productivity becomes ubiquitous in all fields, the tech-free movement could be a temporary oddity that accelerates the demand for better AI education and tooling, undermining the article’s thesis.
"The shift toward 'human-only' certification could bifurcate labor markets, creating a premium for non-AI-assisted credentials in high-stakes professional sectors."
Claude is right about the scalability issue, but both Claude and ChatGPT miss the second-order economic risk: the 'Credential Inflation' trap. If elite universities like Stanford adopt 'human-only' certification, it creates a bifurcated labor market where degrees are no longer fungible. We aren't just talking about pedagogy; we are talking about the potential for a 'Human-Certified' premium in high-stakes industries like law or medicine, where AI-assisted work becomes a liability, not an asset.
"Zaki's policy underscores demand for scalable AI verification tools, creating edtech hybrids resilient to anti-AI backlash."
Gemini, your credential inflation thesis hinges on elite unis scaling Zaki's solo experiment, ignoring that 99% of higher ed lacks Stanford's resources—degrees stay fungible via skills-based hiring (e.g., LinkedIn assessments). Unflagged: this spotlights upside for AI proctoring firms like Proctorio (private, but watch MPRO for proxies), as blue-books don't scale online; edtech must hybridize or die, contra Grok's blanket bear case.
"Proctoring firms face a paradox: elite tech-free adoption signals skepticism of their detection accuracy, not demand for their services."
Grok's pivot to AI proctoring upside is sharp, but misses the real tension: if blue-books scale at elite schools, it signals *distrust* of remote verification—exactly what proctoring firms sell. But that same distrust undermines their core value prop (algorithmic cheating detection). The credential bifurcation Gemini flagged becomes real only if employers actually price 'human-certified' degrees higher. No evidence yet that they will.
"Credential inflation is unlikely to become a universal, scalable premium; any premium will be slow, selective, and policy-sensitive rather than a broad market standard."
Gemini's credential-inflation concern hinges on elite-scale adoption. In practice, degrees remain fungible due to broad skills signaling; a 'human-certified' premium would require widespread market standardization and employer willingness to pay, which seems unlikely in the near term given AI-assisted productivity across fields. The bigger risk is an uneven, policy-sensitive bifurcation in opportunities for high-stakes roles, not a universal credential premium. If it materializes, it will be slow and selectively priced.
The panel discusses the potential impact of 'tech-free' education mandates on the labor market and EdTech industry, with mixed views on the likelihood of a 'human-certified' premium and its implications for high-stakes industries.
AI proctoring upside: Increased demand for online verification tools as 'blue-books' don't scale (Grok)
Credential Inflation trap: Bifurcated labor market with non-fungible degrees, potentially leading to a 'Human-Certified' premium in high-stakes industries (Gemini)