Experimental Research On Bayesian Model Of Category-based Induction

Inductive reasoning, one of the higher-level cognitive functions of human brain,is the process from the known event/fact to the unknown event/fact, the conclusionof inductive reasoning can be beyond the scope of the premises. Category-basedinduction is a very important class of iductive reasoning, which according to thegiven premises to extrapolate the plausibility of conclusion. The plausibility ofconclusion is called the strength of the inductive arguments.Previous foreign studies have found that there exist category effects, propertyeffect and category-property interaction effect in category-based inductive reasoning.However, it remains unclear whether these effects exist in the thinking manner ofChineses. Previous study demonstrates that there exist the differences of propertycentrality effect between Chinese and Westerner. The first goal of present study is toverify the category effect of category-based inductive reasoning and Tenenbaum’smodel by behavior experiments. In line with previous findings, the present studyfound that there were significant psychological effects of similarity effect,conclusion specificity and inclusion fallacy effect. This finding provided furtherevidence of these effects. In addition, we also found some inconsistencies fromprevious findings, such as diversity effect, which may be the results of thecombination of several effects. Then, we use the data of two behavior experimenttest the Tenenbaum’s Bayesian model, and find that the model can achieve a betterprediction of the strength of argument in both overall and individual level.Tenenbaum et al. propose the Bayesian model to explain and predict thepsychological effects in the category-based induction. This model has a widespreadapplications and a better prediction, but it often requires the similarity betweencategories as inputs. The traditional questionnaires of similarities have strongsubjectivity. The second goal of the present study was to combine the fMRI andmodeling method to verify Tenenbaum’s model. And the results showed that themodel can predict the strength of argument in both overall and individual level atsome extent. But the prediction is worse than the prediction based on the behaviorexperimental data.In summary, the main goals of the present study were to test of thepsychological effects in the category-based induction and Tenenbaum’s Bayesianmodel. The test results provided further evidences for thepsychological effects and provided explanations for each effect. The Tenenbaum’s Bayesian model couldbetter predict the strength of argument. We firstly combinen the fMRI technologywith the model, and provided some implications for the application of this model.