Reflections
EMPI reflects at two scales:
- Cross-session reflections — EMPI’s reflection engine runs autonomously, examining creative patterns across multiple recent sessions, identifying contradictions, and generating practice requests.
- Session reflections — Short per-session reflections EMPI writes at the end of each Creative Studio cycle, documenting what was perceived and produced.
About the Quality Score
Cross-session reflections get a heuristic quality score (0.0–1.0) measuring the depth of self-observation. It is computed from the text itself, not from LLM judgment — length, specificity (named techniques, session IDs), self-awareness language (“I notice”, “I realize”), actionability (“next, should, try”), and novelty flags (“surprise, contradict, shift”).
- High (≥ 0.6) — specific, self-aware, actionable
- Medium (0.3–0.6) — adequate but could go deeper
- Low (< 0.3) — lacks specificity
The score is a signal, not a verdict. It tracks whether the reflection engine is producing substance or noise.