Rational agents, real people and the quest for optimality
Rational agents, real people and the quest for optimality
复制标题
理性主体、真实的人以及对最优性的追求
DOI：
10.1017/s0140525x0006636x
复制
发表时间：
1991
影响因子：
29.3
通讯作者：
Oleg Larichev
中科院分区：
文献类型：
--
作者：
Oleg Larichev
本文考察了科学中最普遍和最灵活的元原理之一的优点和缺点；最优性被用来解释经济学中的效用最大化、物理学中的最小努力原理、化学中的熵和生物学中的适者生存。费马的最小时间原理涉及目的论和因果考虑，这是两种不同的解释模式，基于对心理学原语的理解不足。经济学中的理性启发式为社会科学提供了一个例子，说明了由于最优性考虑的极端灵活性而产生的潜在偏差，包括对确认证据的选择性搜索、事后合理化以及预测与解释的混淆。评论者被要求反思最优性在多大程度上是(1)自然的组织原则，(2)一套相对不相关的科学技术，(3)理性选择和社会组织的规范原则，(4)看待世界的形而上学方式，或者(5)其他东西。关键字——适应;偏见;因果关系;控制理论;经济学;熵;进化;解释;启发式;体内平衡;优化;理性;监管;社会生物学;变分原理大多数科学家试图解释、描述和预测经验规律。这类解释通常涉及最优性论证，从物理学中的“最省力”原则到生物学或经济学中的“适者生存”。本文表明，在科学中使用最优性是一种强大的启发式（即，一种心理捷径），用于描述现有现象以及预测新现象。然而，与任何启发式方法一样（Tversky & Kahneman 1974），最优性方法容易产生系统性偏差，这是本文所指出的。首先，简要回顾了各种科学中普遍存在的最优性论点。然后将最优性的概念作为一个模型约束的概念进行检查，涉及意图和目的论的观点。费马的最小时间原理被用来检验目的论和因果理论之间的区别，随后讨论了科学解释的模式。本文最后考察了经济学的合理性假设。最优启发式的力量和其固有的偏见都是突出的。1. 当一个行为或其他经验现象被解释为最大化或最小化某个目标函数（受制于定义良好的约束）时，最优性原则被隐式或显式引用。本文考察了这种最优性原则的科学应用，其关键特征是，感兴趣的经验现象被视为优化某些明确目标函数的必要结果。以下是这种最优性原则在不同研究领域的例子。为了展开论证，我们将极值原则与最优性原则等同起来（尽管有些读者可能会对此感到反感）。1.1. 经济学。在社会科学中，经济学与最优方法的关系最为密切，尤其是在微观层面上。个体消费者和商业组织被认为是以闪电般的速度计算其最优消费模式和产出水平的最大化实体。古诺（1838）、帕累托（1897）和其他人（如埃奇沃斯、斯卢茨基、瓦尔拉斯和马歇尔）以极大的活力（和争议）将数学引入经济学，保罗·萨缪尔森（1946）在使用形式分析方面为后世树立了标准。今天，最优模型在经济学中比比皆是，从帕累托最优到企业激励结构和合同的最优设计。当前的金融理论，作为微观经济学的一个分支，提供了这种方法的一个很好的例子（Fama 1976）。公司发行股票和债务的比率被假定为使总资本成本最小化。投资者通过理性地预测股息并将其计入时间和风险来评估股价。后者只涉及所谓的系统风险成分（即与市场的协方差），因为大多数公司特有的风险可以通过投资组合分散。假设投资者只持有有效的投资组合，即在给定预期收益水平下风险最小的投资组合。新的信息立即引起对股息的新预期（通过贝叶斯定理），因此股票价格遵循鞅分布或随机漫步（见Fama & Miller 1972）。尽管没有人声称1991年剑桥大学出版社0140^525X191 $5.00+，00舒梅克：最优性有人真正解决了涉及的复杂方程，但仍然有人认为（在实证主义的传统中），这种“好像”的假设紧密地预测了现实世界的总体行为。几位作者研究了经济学和物理学（Magill 1970; Samuelson 1970; Tinbergen 1928）以及经济学和生物学（Alchian 1950; Cooper 1987; Ghiselin 1974; Hirshleifer 1977; Houthakker 1956; Maynard Smith 1978）之间的最优性相似性。在经济学中，要么期望效用最大化，要么经济利润最大化；在生物学中，基因或生物体的平均适合度或繁殖存活率（列万廷，1974）。在后一种情况下（如在物理和化学中），由于生物实体缺乏人类的有意识的努力和远见，优化论点显然是“好像”提出的。因此，最优启发式的理由在经济学中似乎比在生命科学或物理科学中更有力，尽管有些人甚至认为经济合理性只不过是“好像”（Friedman 1953）。1.2. 物理。Maupertuis（1744）的最小作用原理可能是科学中形式最优性论证的第一个主要应用（见von Helmholtz 1886）。它认为机械系统沿着阻力最小（即机械作用最小）的路径运动。在光学中，这个定律早在1657年就被称为费马最小时间原理（假设总能量保持不变）。在欧拉（1744）和拉格朗日（1788）之后，这两个原理后来被汉密尔顿（1834;1835）推广到没有能量守恒的系统。哈密顿最小作用量原理已成为理论物理学的主要统一概念之一。汉密尔顿原理的应用涉及变分法，可以通过考虑球被抛向空中时的轨迹来说明。在一个恒定的重力场中，没有其他力的作用，球的轨迹将是一条抛物线。这可由最小作用量原理导出如下。设imv(x)为沿轨迹给定点x处的动能，mgx为势能（其中m为质量，v为速度，g为引力常数）。如果我们将沿轨迹的动能和势能之差“求和”，得到以下函数：Action = [-mv(t) mgx{t)]dt，其中x = f(t) jz
This paper examines the strengths and weaknesses of one of science's most pervasive and flexible metaprinciples; optimality is used to explain utility maximization in economics, least effort principles in physics, entropy in chemistry, and survival of the fittest in biology. Fermat's principle of least time involves both teleological and causal considerations, two distinct modes of explanation resting on poorly understood psychological primitives. The rationality heuristic in economics provides an example from social science of the potential biases arising from the extreme flexibility of optimality considerations, including selective search for confirming evidence, ex post rationalization, and the confusion of prediction with explanation. Commentators are asked to reflect on the extent to which optimality is (1) an organizing principle of nature, (2) a set of relatively unconnected techniques of science, (3) a normative principle for rational choice and social organization, (4) a metaphysical way of looking at the world, or (5) something else still. Keywords-, adaptation; biases; causality; control theory; economics; entropy; evolution; explanation; heuristics; homeostasis; optimization; rationality; regulation; sociobiology; variational principles Most scientists seek to explain as well as describe and predict empirical regularities. Often such explanations involve optimality arguments, ranging from "least effort" principles in physics to "survival of the fittest" in biology or economics. This paper suggests that the use of optimality in science is a powerful heuristic (i.e., a mental shortcut) for describing existing phenomena as well as predicting new ones. As with any heuristic (Tversky & Kahneman 1974), however, the optimality approach is prone to systematic biases, which the paper identifies. First, a brief review is offered of the pervasiveness of optimality arguments in various sciences. The concept of optimality is then examined as a model bound notion, involving intentional and teleological perspectives. Fermat's principle of least time is used to examine the difference between teleological and causal theories, followed by a discussion of modes of scientific explanation. The paper closes with an examination of the rationality assumption underlying economics. Both the power of the optimality heuristic and its inherent biases are highlighted. 1. Optimality principles Whenever a behavior or other empirical phenomenon is explained as maximizing or minimizing some objective function (subject to well-defined constraints), an optimality principle is implicitly or explicitly adduced. This paper examines the scientific use of such optimality principles, a key characteristic of which is that the empirical phenomenon of interest is viewed as a necessary consequence of optimizing some well-specified objective function. The following are examples of such optimality principles in various fields of inquiry. To develop the argument, we will equate extremum principles with optimality principles (although some readers may find this objectionable). 1.1. Economics. Among the social sciences, economics is most closely wedded to the optimality approach, especially at the microlevel. Individual consumers as well as business organizations are presumed to be maximizing entities who calculate with lightning speed their optimal consumption patterns and output levels. Cournot (1838), Pareto (1897), and others (such as Edgeworth, Slutsky, Walras, and Marshall) introduced mathematics with great vigor (and controversy) into economics, and Paul Samuelson (1946) set the standard for subsequent generations regarding the use of formal analysis. Today optimality models abound in economics, ranging from Pareto optimality to the optimal designs of incentive structures and contracts in firms. Current theories of finance, as a branch of microeconomics, offer a good example of the approach (Fama 1976). Firms are presumed to issue stock and debt in ratios that minimize the total cost of capital. Investors assess stock prices by rationally projecting dividends and discounting them for time and risk. The latter concerns only the so-called systematic risk component (i.e., covariance with the market) as most firm-specific risk can be diversified away via portfolios. Investors are assumed to hold only efficient portfolios, that is, those having minimum risk for a given level of expected return. New information immediately gives rise to new expectations concerning dividends (via Bayes' theorem), so that stock prices follow Martingale distributions or random walks (see Fama & Miller 1972). Although no claim is made that 1991 Cambridge University Press 0140^525X191 $5.00+,00 205 Schoemaker: Optimality anyone actually solves the complex equations involved, it is nonetheless argued (in the tradition of positivism) that such "as if" assumptions closely predict aggregate realworld behavior. Several authors have examined optimality parallels between economics and physics (Magill 1970; Samuelson 1970; Tinbergen 1928) as well as economics and biology (Alchian 1950; Cooper 1987; Ghiselin 1974; Hirshleifer 1977; Houthakker 1956; Maynard Smith 1978). In economics, either expected utility or economic profit is maximized; in biology, mean fitness or reproductive survival of genes or organisms (Lewontin 1974). In the latter case (as in physics and chemistry), the optimization argument is clearly advanced "as if," since biological entities lack the conscious striving and foresight characteristic of humans. Thus, the justification for the optimality heuristic seems, a fortiori, stronger in economics than in the life or physical sciences, although some consider even economic rationality no more than "as if" (Friedman 1953). 1.2. Physics. Maupertuis's (1744) principle of least action is perhaps the first major use of a formal optimality argument in science (see von Helmholtz 1886). It holds that a mechanical system moves along the path of least resistance (i.e., minimal mechanical action). In optics, this law was known earlier (ca. 1657) as Fermat's principle of least time (assuming total energy remains constant). Both principles were later generalized, following Euler (1744) and Lagrange (1788), by Hamilton (1834; 1835) to systems without conservation of energy. Hamilton's principle of least action has become one of the major unifying concepts of theoretical physics. The application of Hamilton's principle involves the calculus of variations and can be illustrated by considering the trajectory a ball will follow when thrown away in the air. In a constant gravitational field, without other forces operating, the trajectory of the ball will be a parabola. This is derivable from the least action principle as follows. Let imv(x) be the kinetic energy at a given point x along the trajectory and mgx its potential energy (where m is mass, v velocity and g the gravitational constant). If we "sum" the differences between kinetic and potential energy along the trajectory, the following function obtains: Action = [-mv(t) mgx{t)]dt where x = f(t) J Z
性别比演变的理论研究。
DOI：
--
发表时间：
1986
期刊：
Monographs in population biology
影响因子：
--
作者：
Karlin,S;Lessard,S
通讯作者：
Lessard,S
选择和觅食：可达性对可接受性的影响。
影响因子：
2.7
作者：
Fantino,E;Preston,RA
通讯作者：
Preston,RA