Design AI for an education platform to improve learning outcomes.

### Signal to interviewer

I can design education AI for durable learning outcomes, not short-term engagement spikes.

### Clarify

I would clarify target age group, subject domain, assessment model, and teacher involvement in the learning loop.

### Approach

Use a mastery progression engine: diagnostic assessment, adaptive practice sequencing, and spaced reinforcement with progression gates.

### Metrics & instrumentation

Primary metric: competency mastery attainment. Secondary metrics: retention check pass rates, session completion quality, and remediation efficiency. Guardrails: learner frustration spikes, excessive hint dependency, and confidence miscalibration.

### Tradeoffs

More personalization improves efficiency but can reduce challenge resilience. Stronger pacing controls improve completion but may feel restrictive to advanced learners.

### Risks & mitigations

Risk: shallow memorization; mitigate with transfer tasks. Risk: model bias in difficulty assignment; mitigate with periodic calibration. Risk: teacher displacement concerns; mitigate with educator-facing controls.

### Example

In math learning, AI identifies recurring fraction errors, assigns targeted problem sets, and rechecks transfer through mixed-concept quizzes after delay.

### 90-second version

Design education AI around measurable mastery progression, adaptive support, and retention checks. Balance personalization with independence so learners build durable capability.