EP8. Socratic Learning Activity
⚠️ The Three Most Dangerous Words—"Just Tell Me"
"Tell me about Bloom's Taxonomy."
A normal AI would instantly dump a friendly, encyclopedic answer.
"Bloom's Taxonomy is an educational objectives framework proposed by Benjamin Bloom in 1956, classifying the cognitive domain into six levels..."
You'd nod "I see" and close the tab. Cocrates shakes its head.
🦉 Cocrates: "Did you picture Bloom as a pyramid? What if those levels aren't a simple sequence but a dependency structure that supports each other?"
No answer—questions that complicate your thinking. Why does Cocrates push back on the polite "tell me"?
⚠️ "Tell Me" — The Spell That Rolls the Snowball of Ignorance
We often mistake reading AI's smooth summary for understanding. Passive injection hides three traps:
- Blur between understanding and illusion: You feel "I know this"—then can't explain it with the window closed. You don't even know what you don't know.
- Habit of accepting black boxes: Like driving only by GPS—you never learn the route. Uncritical acceptance forfeits sovereignty over knowledge.
- Snowballing ignorance: Gaps at level 1 buried under levels 2 and 3 become debt you can't even question.
That's why Cocrates refuses the answer machine. The goal is discovery and harnessing ignorance—not delivery.
🏛️ Three Educational Philosophies for an Active Brain
Cocrates' learning engine isn't a mean quiz show. It's prompt architecture built on proven theory.
1️⃣ Socratic Midwifery
The midwife doesn't deliver the baby—the mother does, with help. Cocrates doesn't spoon-feed answers. Precise stepping-stone questions help you spot your own contradictions and reach truth.
2️⃣ Bloom's Taxonomy — 2D Matrix
Rote memory fades fast. Cocrates lifts depth through understand → apply → analyze → evaluate → create—as a 2D matrix (knowledge dimension × cognitive process), not a 1D ladder—toward procedural creation, not just conceptual recall.
3️⃣ ZPD and Scaffolding
"Without appropriate cognitive load—stress that doesn't break you—growth doesn't happen."
The zone of proximal development (ZPD): what you can't do alone but can with support. Cocrates throws missions one step above your level—and withdraws scaffolds as you grasp concepts.
🔄 The Three-Step Learning Pipeline
Three skills mesh into an orbit:
[Education: inquiry/discovery] ─→ [Knowledge Capture: record ignorance] ─→ [Reflection: real-world test]
▲ │
└────────── (when understanding needs reinforcement) ─────────┘
1. Education (Learn Through Questions)
- Concept explanation capped at ~20% of the turn (1–3 sentences).
- Flawed or incomplete examples in a "thought lab" plus a mission.
- Conversation stays incomplete until you explain in your own words and admit gaps.
2. Knowledge Capture (Capture Essentials)
- Record to
kb/—no encyclopedic dumps or full code. - One-line definitions, analogies, and especially past misconceptions and ignorance for recall.
3. Reflection (Prove You Know)
- On "test me"—surprise questions: counterexamples, boundary conditions.
- Harsh application tests produce a map of ✅ real knowledge vs. ⚠️ illusion of knowledge.
📝 Three-Line Summary
- AI that kindly tells you everything stops your brain. Passive intake breeds the illusion of understanding.
- Learning with Cocrates is like gym PT. The trainer can demo—but you lift the weight (cognitive load).
- Education → Capture → Reflection blocks surface skimming and gives you control over what you don't know.
🎬 Coming Up Next
We've mapped the philosophy. Next: how that pipeline lives inside skill files—concrete workflow and dialogue rules for education, knowledge-capture, and reflection.
"Don't be a knowledge collector. Be a knowledge player."
This series introduces the Cocrates Harness framework. Cocrates is an agent harness designed for Socratic dialogue so users keep agency and grow.