Causal And Conceptual Knowledge

描述（由申请人提供）：当前研究项目的目标是了解人们如何学习因果知识和概念知识，以及因果知识和概念知识如何相互作用。特别地，拟议的项目审查了两类对因果归纳有影响的概念。第一类是人们对复杂因果关系结构特征的概念。一个这样的例子是贝叶斯网络中的条件独立假设，它指出，在X导致Y和Y导致Z的因果链中，一旦Y的值已知，X就不能预测Z。在此假设下，上述因果链中X与Z的偶然性成为X与Y之间的偶然性与Y与Z之间的偶然性的乘积。第一个具体目的是检验当人们只面对零碎的协变（如X与Y之间的协变，Y与Z之间的协变）时，是否遵循这个乘积规则，并将它们组合成一条因果链。在因果归纳法中有影响的第二类先验概念是人们对特定事件或物体的知识。我们假设，在共变信息的顺序呈现过程中，人们最初会对学习阶段呈现的特定事件之间的因果关系形成假设，并根据这一初始假设来解释后来的数据。因此，在学习因果关系的过程中，人们更容易受到早期呈现的数据的影响，而不是在同一学习阶段后期呈现的数据，从而产生首因效应。因此，这一建议的一个重要主题是，人们在学习新的因果关系时，在抽象层面（例如，对因果结构施加的约束，而不管特定事件的内容）和在特定层面（例如，关于特定事件的因果效力的概念）都应用了先前的概念。了解因果关系和概念知识具有重要的健康意义，因为外行人和临床医生经常形成疾病及其治疗的因果模型，这些模型极大地影响涉及预防行动和治疗计划的健康相关决策。其目的是超越仅仅演示在因果归纳中使用背景知识，并检查背景概念影响因果归纳的过程的具体性质。

DESCRIPTION (provided by applicant): The objectives of the current research program are to understand how people learn causal and conceptual knowledge, and how causal and conceptual knowledge interact with each other. In particular, the proposed project examines two types of concepts that would be influential in causal induction. The first type is people's concepts about structural characteristics of complex causal relations. One such example is the conditional independence assumption in Bayesian Networks, which states that in a causal chain of X causing Y and Y causing Z, X is not predictive of Z once the value of Y is known. Given this assumption, the contingency between X and Z in the above causal chain becomes the product of the contingency between X and Y and the contingency between Y and Z. The first specific aim is to test whether people follow this product rule when they are presented only with piecemeal covariations (e.g., covariation between X and Y, and covariation between Y and Z) and combine them into a causal chain. The second type of prior concept that would be influential in causal induction is knowledge people have about specific events or objects. It is hypothesized that during sequential presentations of covariation information, people initially form a hypothesis about causal relations between specific events presented during the learning phase and interpret later data in light of this initial hypothesis. Consequently, people would be more influenced by data presented early on during learning of a causal relation than by data presented later in the same learning phase, resulting in a primacy effect. Thus, an overarching theme in this proposal is that people apply prior concepts when learning new causal relations both at an abstract level (e.g., constraints imposed on causal structures regardless of the content of specific events) as well as at a specific level (e.g., concepts about causal efficacy of specific events). Understanding causal and conceptual knowledge has important health implications because laypeople as well as clinicians often form causal models for disorders and their treatments, and these models greatly influence health-related decisions involving preventive actions and treatment plans. The aim is to go beyond mere demonstrations of the use of background knowledge in causal induction and to examine the specific nature of processes in which background concepts influence causal induction.