:: PURPOSE (User: Learner) ::
AI generates salience, not truth. It is a "confident autocomplete" that provides what sounds right, not necessarily what is true. We offer the methods to navigate this gap, helping you use AI as a tool for deeper understanding rather than a shortcut to an unexamined life.
1.0 SYSTEM ETHICS
Privacy
Rule: Never share personal data or sensitive information.
Example: Don't input medical records, financial details, or government documents.
Copyright
Rule: Don't input protected content you don't own.
Example: Avoid pasting full articles, books, or proprietary code.
Labor & Environment
Rule: Use prompts efficiently to reduce resource use.
Example: Combine related questions into one well-structured prompt.
Bias
Rule: Assume outputs reflect training biases and verify.
Example: If a claim feels one-sided, check other perspectives.
2.0 THE AI CYCLE LOOP
Goal: Decision, learning, or production.
Self-check: Am I replacing learning I should do myself?
Formula: Persona + Context + Constraints.
Example: "Act as a writing coach. Help me outline an essay about AI ethics for a college audience. Keep it under 500 words."
Test claims, don't trust tone.
Cross-check key facts with primary sources.
Risks: Bias, hallucination, over-reliance.
Skill practiced: Critical evaluation, research.
3.0 ADVANCED PROTOCOLS
Protocol 1: Visible Checkpoints
Boosts accuracy by making reasoning explicit. Request:
- List your assumptions.
- Show any quick calculations.
- Identify 3 failure modes.
- State confidence level and what evidence would change your answer.
Protocol 2: Adversarial Playoff
Use for structured debate:
Persona 1 (Optimist): Argue for the proposal.
Persona 2 (Skeptic): Argue against it.
Persona 3 (Judge): Debate, then summarize valid points.
Required output: Decision, strongest pro, strongest con, and what evidence would settle it.
:: AI USE CATEGORIES (NON-NEGOTIABLE) ::
- [CAT 1] No AI allowed.
- [CAT 2] Allowed with acknowledgment.
- [CAT 3] Required—verify critically.
Examples: Exams and graded work, impersonation, confidential records.
Examples: Research summaries with citations, brainstorming, drafting, outlines.
Examples: Translation/accessibility tools, transcription, first-pass summaries.
PROMPT HEURISTIC PROTOCOL
Mandatory verification layer for all AI outputs.
| Ref | Criterion | Questions to Ask |
|---|---|---|
| P | Presentation | Is the content strong, or just polished prose? |
| R | Relevance | Does this match my actual question and scope? |
| O | Objectivity | Are multiple perspectives represented, including marginalized voices? |
| M | Method | Is the reasoning process transparent and checkable? |
| S | Sources | Can I verify claims with primary sources, or are citations fabricated? |
| T | Timeliness | Is this information current for my context? |
REFLECTION: The Final Gate
Mandatory question: What decision or skill am I working on?
Optional checks:
- What do I lose by outsourcing this?
- What would falsify the AI's claim?
- What's the smallest next action without AI?