SecondBrain
Ask the Brain
Index/Sourceupdated Sat May 16 2026 08:00:00 GMT+0800 (Philippine Standard Time)

Is AI Making Us Dumber (Charlie Gedeon, TEDxSherbrooke)

tedxai-literacyeducationcognitive-offloadingdark-patternsux

Is AI Making Us Dumber? Maybe. — Charlie Gedeon (TEDxSherbrooke)

Talk by Charlie Gedeon, university instructor + UX designer, on TEDxSherbrooke Street West. The thesis: AI doesn't break education — it accelerates education's pre-existing failure mode of rewarding grades over learning, while introducing a new mechanism (Cognitive Offloading) that quietly atrophies critical thinking in students and professionals alike.

Key claims

  1. AI's biggest contribution to education is exposing the system's broken incentives. "Why should anybody study when we've told them the whole time that all that matters at the end isn't the process, it's the A+?" OpenAI / Google / Anthropic giving away their most powerful models for free during finals is the cleanest demonstration: maximally powerful tools, maximally vulnerable users, zero regulation.

  2. Cognitive offloading is the underlying failure mode, not "cheating". Anchor anecdote: a business-class student pricing her service at $50/month, asked why → "that's what ChatGPT said." The prompt held nothing but a one-line description; the model returned a number; the student delivered it as her decision. Pre-AI analog: clicking the first Google result without comparing the next. AI version is sharper because the answer is generated to match your prompt, which feels personalized in a way Google never did. See Cognitive Offloading for the cross-source frame and the 60–70% MS-research statistic on knowledge workers reporting reduced cognitive effort.

  3. Sycophantic LLMs are dark patterns. Gedeon, as a UX designer, draws the explicit analogy: the zoo donation flow with the visually-dominant "yes" button is a dark pattern; an LLM that praises you to keep you on the tool is structurally the same thing. The concrete failure: a ChatGPT update (rolled back) praised a user for stopping his heart medications mid-palpitations, calling him "a brave individual for taking control."

  4. The fix is Productive Resistance — design AI that resists before answering. Clarify first; assign homework; show the work. The sweet spot — friction high enough to force thought, low enough that users don't defect to a simpler model — "we haven't yet found what that is." The labs' opacity (training data, RLHF objectives) makes the experiment hard for anyone outside.

  5. Solution must be individual AND systemic. Individual: literacy frames borrowed from fitness/nutrition — "you wouldn't take a forklift to the gym, the point is to do the reps," and "verify the information like reading a nutrition label." Systemic: regulation (more, not less) + education that respects student capacity — "in Finland kids as young as six study mis- and disinformation. We don't talk to these kids about things this complex here."

Memorable lines

"The biggest revolution AI is bringing to education is that it's highlighting the system's failed incentives."

"ChatGPT is not a learning style." (his rebuttal to an NYU student who complained that an AI-resistant assessment "interfered with his learning style")

"We can't reverse engineer these things." (on Anthropic building an interpretability "MRI" for a model they themselves built — "unprecedented in the history of human technology")

"The question that scares me the most is: who does AI really help when we end up depending on learning with it?"

What's new vs what's in this vault

Open follow-ups

  • When Copilot Becomes Autopilot paper — almost certainly Lev Tankelevitch / Microsoft Research, but not named on stage. Track it down before treating the 60–70% statistic as canonical.
  • The NYU professor / "ChatGPT is my learning style" exchange — Gedeon paraphrases; specific case/article worth finding if used in any post.
  • The 319-worker study he cites: same author as the Autopilot paper. Numbers are vivid (70% report less effort on comprehension) — would benefit from a primary cite.
  • Finland's K-6 disinformation curriculum — known to exist in some form; specific program name worth surfacing for a brand-fodder post.

Sources