---
aliases: {}
concept_id: human_values
preferred:
  en: human values
  ja: 人間の価値観
relations:
  relations:
    broader:
    - values
status: active
---

# human values

The evaluative commitments, ends, and features of outcomes that humans care about or regard as good, important, or worth preserving; in AI alignment the phrase often denotes a difficult-to-specify target rather than a single agreed formal objective.

## broader

- [values](../terminology/values.txt)

## Documents: mentions

- [「Alignment by Default」への強い懐疑](../documents/akira.alignment-by-default-skepticism.txt)
- [望ましい振る舞いは人間を重視する目標の証拠ではない](../documents/akira.behavioral-alignment-is-not-goal-alignment.txt)
- [知能水準と最終目標は、原理上ほぼ独立である](../documents/bostrom.orthogonality-separates-intelligence-from-final-goals.txt)
- [能力はアラインメントより遠くまで汎化しうる](../documents/yudkowsky.capabilities-generalize-beyond-alignment.txt)
- [能力と目的は別の軸である](../documents/yudkowsky.capability-motive-separation.txt)
- [CEV 型 Sovereign と訂正可能な AGI は異なる安全化戦略である](../documents/yudkowsky.cev-sovereign-vs-corrigible-agi.txt)
- [人間にとって価値ある未来は、価値体系の一部を欠くだけでも大きく損なわれ得る](../documents/yudkowsky.value-fragility.txt)

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