Heuristics and Biases are (Nearly) Optimal: A Fresh Programmatic Study of Heuristics and Biases in Human Decision-Making

我们的决定塑造我们的生活：选择什么样的职业道路；如何投资；选择合作伙伴；如何获得幸福。我们的应用程序提出了计算机模拟、数学建模和心理学实验来理解和解释人类决策的核心问题。我们研究项目的总体目标是通过系统地为人类决策中的启发式和偏见建立理性基础，提供一种理解它们的新方法。我们的研究项目挑战了人们普遍认为的人类认知是系统性错误的观点，通过证明在给定环境条件和认知限制的情况下，决策几乎是最佳的，从而提供了一种新的、积极的认知观点。普遍的观点是，判断和决策充满了认知偏差和理性偏差，因为人们使用快速但不准确（次优）的启发式。我们将为启发式建立数学上严格的最优性保证，强烈挑战这一主流观点，从而提供一个完全不同的启发式观点。我们的新框架补充了理性研究，通过试图表明启发式在给定环境条件和认知限制下几乎是最佳的，将其发展成为一个更丰富的研究项目。我们的初步工作为这一系列工作提供了支持证据，表明我们的框架为众所周知的启发式“尽力而为”提供了强有力的最佳保证。此外，与获得诺贝尔奖的前景理论和其他现有方法相比，我们提出了一种新的启发式决策模型，称为SbEU（基于样本的期望效用）。SbEU有效地结合了两种非常成功的认知理论，期望效用理论和有限最优，独特地模拟并解释了人类决策中的广泛重大发现。值得注意的是，SbEU是第一个，也是迄今为止唯一一个在风险、基于价值和博弈论决策之间架起桥梁的有界最优启发式模型。我们的研究计划提供了一个新颖的，统一的，精确的，正式的解释了一些关于人类决策的大量文献。我们的初步模型将被改进、扩展，并应用于更多的实证研究结果。我们的模型的新颖预测将在新的心理学实验中得到实证检验。我们的假设是，人类以一种有界最优的方式做出决策：根据环境条件以及他们所面临的计算和认知限制做出理性决策。我们希望这项工作能引发认知科学的重大范式转变，从看似无穷无尽的无法解释的偏见到有限理性的统一理论。这将从根本上改变我们目前理解人类认知的方式，对决策科学、政策制定和日常生活决策产生深远影响。因此，我们的研究对科学家和外行人都很重要。

Our decisions shape our lives: what career path to choose; how to invest; choice of partner; how to achieve happiness. Our application proposes computer simulations, math modeling, and psychology experiments to understand and explain central issues in human decision-making. The broad goal of our research program is to provide a new way of understanding heuristics and biases in human decision-making, by systematically developing a rational basis for them. Our research program challenges the widely held view that human cognition is systematically erroneous, offering instead a new and positive view of cognition by demonstrating that decision-making is nearly optimal given environmental conditions and cognitive limitations. The prevailing view is that judgment and decision-making are filled with cognitive biases and deviations from rationality, because people use fast but inaccurate (sub-optimal) heuristics. We will establish mathematically rigorous optimality guarantees for heuristics that strongly challenge this dominant view, thus offering a radically different perspective on heuristics. Our new framework complements rationality research, developing it into a richer research program by attempting to show that heuristics are nearly optimal given environmental conditions and cognitive limitations. Providing supporting evidence for this line of work, our preliminary work shows that our framework yields strong optimally guarantees for a well-known heuristic, Take-the-Best. Also, in contrast to Nobel-winning Prospect Theory and other current approaches, we present a new heuristic model of decision-making, called SbEU (Sample-based Expected Utility). Effectively combining two highly successful theories of cognition, expected utility theory and bounded optimality, SbEU uniquely simulates and thus explains a wide range of major findings in human decision-making. Notably, SbEU is the first, and so far the only, boundedly-optimal heuristic model that bridges between risky, value-based, and game-theoretic decision-making. Our research program provides a novel, unified, precise, formal explanation of several vast literatures on human decision-making. Our preliminary models will be refined, extended, and applied to more empirical findings. Novel predictions of our models will be empirically tested in new psychology experiments. Our hypothesis is that humans make decisions in a boundedly-optimal fashion: making rational decisions, subject to environmental conditions and the computational and cognitive limitations they are faced with. We expect this work to provoke a major paradigm shift in cognitive science, from a seemingly endless list of unexplained biases to a unified theory of bounded rationality. This would fundamentally change the way we currently understand human cognition, with far-reaching implications for decision sciences, policymaking, and daily-life decisions. As such, our research is of great importance for scientists and laypeople alike.