Artificial Intelligence and the Work of High Quality Learning by Jason Lodge HIGHLIGHTS

Main ideas, concepts, and lessons

• Focus on the “human side” of AI in learning

  • AI tools affect humans and human learning, especially through human–machine interaction.
  • There is skepticism: current AI tools (especially generative AI) were not built for learning, but for other goals like productivity and marketing. Learning use therefore requires adjusting how we use them and understanding potential downsides.

• Effort in learning matters (AI can reduce needed effort)

  • Analogy: before e-bikes, riding required pedaling and built fitness; with e-bikes, you can go without pedaling and may not get fit.
  • Parallel: before generative AI, learning required effort—effort helped the brain build skill.
  • Risk: with generative AI, learners may not “pedal,” so the effort that builds capability can disappear.
  • Lesson: use AI in ways that preserve meaningful cognitive effort.

• Generative AI differs from a calculator

  • Calculator
    • Assumes you already understand domain symbols/operations (needs prior knowledge).
    • Is mostly transactional: input → output → move on.
    • Does not guide you while you’re using it.
  • Generative AI
    • Can be used with less domain knowledge (lower barrier to entry).
    • Often provides instructions/explanations while engaging with it.
    • Therefore it changes how learning happens, not just how answers are produced.

• Three broad ways to use technology in learning

  1. Offload (individual level; tool-like, transactional)
     • Learner provides the task/inputs to the machine and receives an answer.
     • Example: using AI like a calculator—submit, get output, move on.
  2. Extend (individual level; tool-like expansion of capability)
     • Technology increases the learner’s capacity to do something.
     • Key distinction from offloading: it’s about adding capability, not just replacing effort.
  3. Co-regulate / Collaborate (peer-like interaction; machine guides learning)
     • Learner works with the machine to complete the task cooperatively.
     • Machine helps guide decisions and progress (co-regulation).

• How learners actually use AI in context

  • Students often start with AI when stuck because it provides clear explanations.
  • Students also know AI can be wrong, so they check the internet to verify.
  • Learners value non-text media (video, podcasts, multimedia) for deeper understanding; generative AI is often text-based, which can limit visual/multimedia learning unless supplemented.
  • Some students eventually prefer interacting with a human, saying AI interaction can feel unsatisfying or unhelpful, and they need humans to test knowledge.

• Best results come from being a good learner first

  • Students’ success with AI is not about understanding the machine best.
  • The crucial factor is: good learners first, then they can integrate AI appropriately among other learning resources.

• Learning that matters is hard

  • Education is compared to going to a personal trainer, not paying for entertainment like Netflix.
  • “Greatest skill in the age of AI”: being a good self-regulated learner (from the World Economic Forum report).
  • Hard mental work feels unpleasant, and many people avoid difficulty.
  • Lesson about confusion:
    • Feeling confused does not mean you aren’t good enough.
    • Confusion indicates investment and is often a prerequisite for the “aha moment.”
    • You don’t get insight without first being confused.

• What self-regulated learning involves (including metacognition)

  • Two parallel processes during learning:
    1. Task-level learning: trying to understand and complete the learning goal.
    2. Monitoring + decisions: judging progress and choosing where to go next.
  • The second layer is metacognition (thinking about one’s own understanding).
  • Example of miscalibrated self-assessment:
    • Someone reads a highly structured Wikipedia page on volcanoes and concludes they are an expert—illustrating how learning progress can be misaligned with perceived understanding.
  • Expert learners know what they don’t know and can judge their knowledge accurately.

• Limits of AI related to metacognition and evaluation

  • AI explanations may be liked, but AI is weaker at:
    • Evaluating and making sense of things (inference).
    • Regulating itself, because it lacks conscious metacognition.
  • Human metacognition is presented as a key “superpower.”

---

Methodology / instruction-style guidance (as described)

• Use AI in ways that do not remove productive effort

  • Aim for AI use that supports real thinking (avoid becoming a passive “rider” who doesn’t “pedal”).
  • Treat AI as a support for learning progress, not a shortcut that bypasses understanding.

• Prefer self-regulated, effortful learning over answer offloading

  • When stuck, consult AI—but pair it with verification and other learning sources.
  • After AI explanation, check reliability (e.g., consult the internet).
  • Use multimedia resources (video/podcasts) when they improve comprehension.

• Practice the two-process cycle of self-regulation

  • While learning:
    • Process 1: Do the learning work (understand concepts, complete tasks).
    • Process 2: Continuously monitor progress and decide next steps.
  • If perceived understanding may be wrong (misalignment), refine your judgment—good learners recognize what they don’t know.

• Interpret confusion as a sign of engagement

  • Do not treat confusion as evidence of inadequacy.
  • Use confusion as a signal you’re close to deeper understanding and might be on the path to an “aha moment.”

• Match the AI “mode” to the learning goal

  • Offload only when appropriate (answer retrieval may undermine learning effort).
  • Extend when you want capacity-building rather than replacement.
  • Co-regulate/collaborate when you want guidance that helps you make decisions during learning.

---

Speakers / sources featured

• Speaker: Jason Lodge (implied by the video title; he delivers the remarks)
• Source cited: World Economic Forum (report mentioned, about self-regulated learning as the key skill in the age of AI)
• Other speakers: No other specific speakers named in the provided subtitles.