A general instance-based learning framework for studying intuitive decision-making in a cognitive architecture

Cognitive architectures (e.g., ACT-R) have not traditionally been used to understand intuitive decision-making; instead, models tend to be designed with the intuitions of their modelers already hardcoded in the decision process. This is due in part to a fuzzy boundary between automatic and deliberative processes within the architecture. We argue that instance-based learning satisfies the conditions for intuitive decision-making described inKahneman and Klein (2009), separates automatic from deliberative processes, and provides a general mechanism for the study of intuitive decision-making. To better understand the role of the environment in decision-making, we describe biases as arising from three sources: the mechanisms and limitations of the human cognitive architecture, the information structure in the task environment, and the use of heuristics and strategies to adapt performance to the dual constraints of cognition and environment. A unified decision-making model performing multiple complex reasoning tasks is described according to this framework.

认知架构（例如ACT - R）传统上并未被用于理解直觉决策；相反，模型往往是在其建模者的直觉已经在决策过程中被硬编码的情况下设计的。这部分是由于架构内自动过程和审慎过程之间的模糊界限。我们认为，基于实例的学习满足了卡尼曼和克莱因（2009）所描述的直觉决策的条件，将自动过程与审慎过程区分开来，并为直觉决策的研究提供了一种通用机制。为了更好地理解环境在决策中的作用，我们将偏差描述为源于三个方面：人类认知架构的机制和局限性、任务环境中的信息结构，以及为使表现适应认知和环境的双重约束而使用的启发式方法和策略。根据这一框架，描述了一个执行多种复杂推理任务的统一决策模型。