关键词: 类比推理, 因果模型, 计算模型

Abstract: Analogical inference is important for academic tasks and daily life. There are two kinds of extant models, computational models and causal models, trying to explain the analogy process in different ways. Computational models of analogy assume that the strength of an inductive inference about the target is based directly on similarity of the analogs. In contrast, causal models suggest that analogical inference is also guided by causal models of the source and target. Lee and Holyoak (2008) reported that analogical inference appeared to be mediated by building and running a causal model. However, the causal model adopted in their materials was common-effect model, which was merely one kind of causal models. Causal models include cause-effect and effect-cause in the reasoning directions; unique cause-effect model, common-effect model, common-cause model, and multiple cause-effect model in the feature dimensions. More importantly, the common-effect model is significantly different from the common-cause model both in reasoning direction and reasoning difficulty. Then whether the results of Lee and Holyoak (2008) can represent all kinds of causal models? To answer this question, this research focused on the possibility that people make analogical inference by running causal models in cause-effect and effect-cause order, under the conditions of common-effect and common-cause feature structure.
Two experiments were performed to investigate the possibility that people make analogical inference by running causal models. About fifty undergraduates were randomly selected to participate in each experiment. Participants were asked to read a description of imaginary animals in a booklet, and then evaluate the inductive strength of analogical inference. In experiment 1, participants were asked to make analogical inference in a cause-effect order under the conditions of common-effect and common-cause feature structure. In experiment 2, participants were asked to make analogical inference in an effect-cause order under the conditions of common-effect and common-cause feature structure.
The results of experiment 1 showed that in cause-effect direction, participants used causal models to make analogy in common-effect feature structure, but used both causal models and computational models in common-cause feature structure. The results of experiment 2 showed that in effect-cause direction, participants used causal models to make analogy in both common-effect and common-cause feature structure.
The present findings indicated that people could build and run causal models in both cause-effect order and effect-cause order to make analogical inference. When computational models and causal models competed with each other, people tended to use causal models in analogical inference; but if the two models didn’t compete, people tended to use both. Future work should be focused on building a more perfect model to explain the cognitive process of analogical inference.

Key words: analogical inference, causal models, computational models