Testing and extending a dual-source model of everyday conditional reasoning

根据概率条件推理的双源模型，人类推理利用两个信息源：给定前提的句法形式和特定于内容的背景知识。我们假设模型的基于形式的部分是由(资源有限和容易出错的)单调推理系统驱动的，而基于知识的部分是由对长期和暂时突出的信念的摘要评估驱动的，违反了不同背景下的一致性和单调性。该模型被应用于三组经典问题：1)语境效应，特别是抑制效应和说话人专业知识的影响。抑制效应经常被视为非单调人类推理的证据，因此应该映射到我们模型的基于知识的部分，而说话人专业知识表面上类似的影响应该映射到基于形式的证据和基于知识的证据的相对权重上。2)该模型将被扩展到“IF-THEN”以外的其他连词，以观察不同的句法形式是否以与心理逻辑的存在相一致的方式映射到基于形式的成分上。3)评估认知资源对模型中基于知识和基于形式的部分的相对作用。

According to the dual source model of probabilistic conditional inference, human inferences draw on two sources of information: The syntactical form of given premises and content-specific background knowledge. We assume that the form-based component of the model is driven by a (resource-limited and fallible) monotonic system of reasoning, whereas the knowledge-based component is driven by summary assessments of chronically and momentarily salient beliefs, violating, across different contexts, coherence and monotonicity. The model is to be applied to three groups of classical problems: 1) Context effects, in particular suppression effects and effects of speaker expertise. Suppression effects are frequently seen as evidence for non-monotonic human reasoning and should thus map on the knowledge-based component of our model, whereas superficially similar effects of speaker expertise should map on the relative weight of form-based versus knowledgebased evidence. 2) The model is to be extended to other conjunctions than “if-then” to see whether the different syntactical forms are mapped on the form-based component in a manner consistent with the existence of a mental logic. 3) The relative role of cognitive resources for the knowledge-based and the form-based components of the model is to be assessed.