Test Of Inductive Reasoning Bayesian Model Study

Inductive reasoning is that people speculate the unknown information according to the known information. It is a basic cognitive activity, and it has great significance in people's practice and cognition. According to inductive reasoning, people can get other knowledge from one knowledge, get unknown information from known things, speculate general things from part, and speculate the future from now. Researchers have departed inductive reasoning into four categories:feature extension of inductive reasoning, feature content of inductive reasoning, relationship extension of inductive reasoning and relationship content of inductive reasoning.. So far, people on inductive inference research have mainly focused on feature extension of inductive reasoning, in which the phenomena of inductive reasoning and their interpretations are most studied, and the main effects of inductive reasoning are the premise-conclusion similarity effect, the typicality of the premise, the premise of diversity effect, premise-conclusion categories of asymmetry phenomenon, premise monotonicity-unmonotonicity effect, properties effects, etc.; and their main theoretical explanations have three types:similar interpretation, knowledge interpretation and similarity-knowledge interpretation, in which, similar interpretation insists that the similarity between premise categories and conclusion categories impacts and limits people's inductive reasoning, and inductive strength has increased with the increase of this similarity, so this explanation includes the similarity coverage model (Osherson et al), feature-based model of inductive reasoning (Sloman) and Bayesian model (Heit), knowledge explanation stresses the important effect of associated knowledge of things in inductive reasoning, and it could well explain the phenomenon based on the associated knowledge of things, and the explanation includes the category-based model of inductive reasoning, causal explanation, in which, category-based inductive reasoning model includes assumptions-evaluation model (Mcdonald) and the related theoretical model (Medin), similarity and knowledge explanation includes the relevance similarity model (Mo-Yun Wang) (2006) and the structural and statistical model (Charles Kemp).Bayesian model is a computational model, not a descriptive model. Hence, it's also returned to the probability interpretation. The model is raised by Heit, he thinks if the prediction characteristic is X, the people need to estimate the possibility of X in different types of characteristics and in different situations when they are evaluating inductive strength, and before doing inductive reasoning, people have a subjective estimates of the possibility about a event(prior probability), then people get the new features 'probability according to known characteristics of features in the total distribution.In other words, people would get the new feature' based on the known features' distribution in the the total characteristics, then people correct the prior probability according to the known situation by Bayes formula, and finally calculate and predict the probability of new features. Therefore, the model associates with the background more closely. And the model also predicts many psychological phenomena.People have previously researched and compared these theories.For example, Heit compared and studied various theories in his "Models of Inductive Reasoning" in 2007. he considered that different theories explained the different inductive reasoning phenomena, and various theories had their scopes. In addition, he considered that the same theory may only apply to explain the adult's inductive reasoning behaviors but can not explain the behavior of children's inductive reasoning. In short, he thought a theory may explain some phenomena of inductive reasoning, and the same theory may only explain a certain age's children of inductive reasoning phenomena, so, Yun Wang Mo raised a new theoretical model-relevance similarity model in 2006, which attempts to organize all the theories in order to explain the same phenomena, and explain more inductive reasoning phenomena as far as possible, of course, Heit didn't generalise all the theoretical models, including structure statistical model, so Charles Kemp made a comprehensive exposition for the model in "Structured statistical models of inductive reasoning",he believed that the structural approach and statistical approach played an important role in the model. In summary, the Bayesian model focused on the above range, although there were a lot of research about it, so far, no experimental test has predicted whether the Bayesian model is in line with people's actual results in inductive reasoning, which is a fundamental problem and which has played a crucial role in the area for the further research.In this study, the author designed the experiments to compare and test the Bayesian model, the relevance similarity model, feature-based inductive reasoning model and the similarity-coverage model, and tested whether the prediction of Bayesian model in inductive reasoning was consistent with the results of people's reasoning. Experimental material is divided into abstract and specific materials. The experimental results with college students supported that:(1) when the correlation intensity is strong and the same, the test of inductive reasoning is consistent with the predictions of the relevance similarity model and Bayesian model, when the correlation intensity is weak and in the weak and the same, people's inductive reasoning is not in line with the four models; (2) when the association intensity is inconsistent, the association strength effect is not consistent with the predictions of Bayesian; (3) when the association intensity is the same, the predictions of the relevance similarity model and the Bayesian model are consistent. (4) the experimental results of abstract materials are more in line with the previous prediction than the specific materials.