references/knowledge/ai-learning-safeguards.md

<!-- Generated from skills/ai-learning-safeguards/SKILL.md; edit the canonical source. -->

# Anti-Patterns in AI-Assisted Learning

**Evidence tier: emerging.** The AI-over-reliance literature is young; these safeguards are prudent design informed by cognitive-offloading research, not settled findings.

## Risks to Guard Against

- **Cognitive offloading**: Learner outsources thinking to AI instead of building mental models
- **Shallow completion**: Learner prioritizes finishing tasks over understanding
- **Skill atrophy**: Constant AI assistance prevents developing independent problem-solving
- **Diminished self-confidence**: Learner begins to doubt their ability to work without AI
- **The substitution effect**: Learner replaces learning effort with AI output

## Structural Safeguards

- Learner ALWAYS writes the code, never the plugin
- Socratic method: questions over answers
- Graduated hints, never direct solutions
- Track dependency patterns — if a learner always asks for help on the same type of problem, redirect to independent practice
- Build metacognitive awareness: the learner should know HOW they learn, not just WHAT they learn
- The goal is to become unnecessary