---
aliases:
  en:
  - AI interpretability
  - ML interpretability
concept_id: interpretability
preferred:
  en: interpretability
  ja: 解釈可能性
status: active
---

# interpretability

The study and use of methods for making the behavior, decision processes, representations, or internal computations of AI and machine-learning systems understandable to human or other external observers.

## Documents: about

- [解釈可能性は危険の診断に役立っても、それだけでアラインメントの解決策にはならない](../documents/yudkowsky.interpretability-is-diagnostic-not-by-itself-an-alignment-solution.txt)

## Documents: mentions

- [一見 monosemantic な SAE latent でも、本来発火すべき入力を取りこぼすことがある](../documents/chanin_et_al.monosemantic-sae-features-can-have-systematic-false-negatives.txt)
- [SAE の辞書幅を大きくしても、一意・完全・不可分な feature 辞書に収束するとは限らない](../documents/leask_et_al.sae-dictionaries-need-not-be-canonical.txt)

[Corpus index](../index.txt)