Herbert A. Simon

Herbert A. Simon：诺贝尔经济学奖/图灵奖得主，层级系统与近可分解性理论，AI 先驱

实

ENTITY · HERBERT A. SIMON · NOBEL 1978 · TURING 1975 · BOUNDED RATIONALITY · ARCHITECTURE OF COMPLEXITY

Herbert A. Simon

Thinker spanning economics, cognitive science, and AI (1916–2001) — founder of bounded rationality and complex-systems theory

Simon was one of the rare thinkers who truly crossed disciplinary lines — his work stretches across management, economics, cognitive science, AI, and complex-systems theory. 1978 Nobel Prize in Economics (bounded rationality), 1975 Turing Award (shared with Newell, for contributions to AI and cognitive psychology). Bounded rationality is deeply isomorphic to the constraints LLM agents face: every decision made in a finite context window is an instance of bounded rationality.

Core contributions

Architecture of Complexity (1962)Hierarchical systems + near-decomposability + the watchmaker parable + description compressibility — a general theory of complex systems

PSSH (1976, with Newell)A physical symbol system is a necessary and sufficient condition for general intelligent behavior — the founding claim of symbolic AI

Bounded rationalityDecision-makers’ rationality is bounded by information, cognitive capacity, and time — optimization gives way to satisficing

Connections to AI agent engineering

Near-decomposability → modularity

Intra-level interactions dominate inter-level ones — the theoretical basis for modular design of agent systems

Bounded rationality → context limits

LLM agents deciding inside a finite context window are bounded rationality realized in a computational substrate

Contrast with Aristotle

Simon decomposes complex systems by structure; Aristotle by explanatory role — complementary frames for analyzing complexity

→ Hierarchical Systems · Near-Decomposability · Physical Symbol SystemSimon (1962) / Newell & Simon (1976)

Herbert A. Simon

简介

Herbert Alexander Simon（1916-2001），美国政治科学家、经济学家、认知心理学家、计算机科学家。1978 年诺贝尔经济学奖得主（有限理性理论），图灵奖得主（1975，人工智能与认知心理学贡献）。

Simon 是少数真正跨越学科边界的思想家——他的研究横跨管理学、经济学、认知科学、人工智能和复杂系统理论。

与本 wiki 的关联

The Architecture of Complexity（1962）

Simon 最具持久影响力的论文之一。笔记摘要

论文提出的核心框架：

层级系统：复杂系统普遍呈现层级结构
近可分解性：层级系统内部交互强于层级之间的交互
钟表匠寓言：稳定中间形态加速复杂性的演化涌现
描述的可压缩性：正确的表示方式可以大幅简化对复杂系统的描述

人工智能先驱与物理符号系统假说（PSSH）

Simon 与 Allen Newell 合作开发了 Logic Theorist（1956）和 General Problem Solver，是符号 AI 的奠基人之一。论文中”问题求解即搜索”的框架直接来自这些工作。

1976 年图灵奖讲座提出计算机科学最基本的定性结构律——物理符号系统假说：

“一个物理符号系统具备通用智能行为的充分必要条件。”

配套提出启发式搜索假说：物理符号系统通过在问题空间中生成并递进修改符号结构来求解问题，核心技术是手段-目的分析。

有限理性（Bounded Rationality）

Simon 提出的有限理性概念——决策者的理性受限于可获得的信息、认知能力和时间约束——与当代 agent 系统面临的约束高度同构。LLM agent 在有限 context window 中做出的每个决策，都是有限理性的实例。

合作者

论文致谢中提到的关键人物：

Allen Newell：认知科学与 AI 领域的长期合作伙伴
Warren Weaver：复杂性科学先驱，区分了”有组织的复杂性”与”无组织的复杂性”

References

Simon, H. A. (1962). “The Architecture of Complexity”. Proceedings of the American Philosophical Society, 106(6), 467-482. 笔记: sources/simon-architecture-of-complexity-notes.md
Newell, A., & Simon, H. A. (1976). “Computer Science as Empirical Inquiry: Symbols and Search.” Communications of the ACM, 19(3), 113–126. 摘要: sources/newell-simon-computer-science-empirical-inquiry-1975.md