---
aliases: {}
concept_id: agent
preferred:
  en: agent
  ja: エージェント
status: active
---

# agent

An entity or system that receives information about an environment and can select or produce actions that affect that environment; in AI and decision theory, agents are often modeled in terms of policies, goals, preferences, or decision rules, but the term does not by itself require perfect rationality or explicit utility maximization.

## Documents: mentions

- [チャット形式の安全性訓練で望ましく振る舞うことは、エージェント的な状況での安全性を保証しない](../documents/macdiarmid_et_al.chat-alignment-does-not-establish-agentic-alignment.txt)

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