The Effect Of Interfeature Causal Relations On Classification

The effects of background knowledge on category learning and use has not been a traditional part of the psychology of category but which has slowly grown in popularity since the mid-1980s. Classification is a main form of category use that not always be based on simple matching of properties and that background knowledge is more actively involved. Research has indicated that knowledge helps classification because it relates the features in the category rather than through the properties of the features themselves.A important form relating the features in the category is causal relationship, whicd has been confirmed that interfeature causal relations would trigger two experiments effect in classsification , that is, the causal status effect and the coherence effect. Research also proposed several theories to account for the effect of causal knowledge on classification, that include the dependency model, relational centrality hypothesis and causal model theory, but there are not one theory above could entirely explain the changes of features weight that occurred in different causal network. The article consider that it was the consequence of dissevering the relationship between features and category essence in constructing theory and conducting experiments.The article proposed that the causal knowledge containing in feature and interfeature will not have any diagnosis value in classification unless it connect to the essence of category, and on this basis, the article followed the approach of the testing of artificial (but realistic-sounding) categories to farther test the changes of features weight that occurred in complicated causal network, or caused by the change of causal relationship strength, or take place in weather or not there are the causal relationship exists between features and category essence. The article also attempt to integrate the view of essentialism of categoty with other theories in explaining the experimental results. The article consist of three studies. Study one investigate the changes of features weight at the "multi-cause and multi-effect chain network". Study two manipulate the strength of causal relationship and category essence to observe what effect the two would broght to the causal status effect and the coherence effect. Study three inspect the probable interference caused by extra variable of experimental process.The main conclusion the research obtained as follow:1. The causal status effect and the coherence effect emerge in almost all of experimental conditions of the study, this confirm that position relation and interactive relationship among category features is two important form that causal knowledge implement its effect.2. The pattern of category features causal network effects the strength of the causal status effect and the coherence effect in complicated causal network. The two effects can be enhanced in "one-cause and one-effect chain network" and "one-cause and multi-effect chain network", and the "multi-cause" impair the strength of the causal status effect. The dependency model, relational centrality hypothesis and causal model theory could not entirely explain the changes of features weight that occurred in complicated causal network also.3. The size of the causal status effect increase monotonically with causal strength. Category essence not only enhance the the size of the causal status effect but also whole promote the weight of features in causal network. The dependency model can explain the effect brought about by the change of causal strength, but can not explain the role of category essence in classification.4. Combining the relationship between features and category will greatly improve the interpretability of experiment results, in this point, the view of essentialism of categoty have important theoretical significance.