A Study About The Influence Of Working Memory On Bayesian Reasoning

In our daily life, we always had to make some life-and-death decisions on the basis of conditional probabilities, such as in context as diverse as the court room, the doctor’s office and the war room. Although these decisions or judgement were important, people often did not perform so good as expect. One typical example was Bayesian reasoning.After Edwards(1968) firstly introduced this question,studies had shown that people were not wise decision-makers,they performed poorly in such kind of Bayesian tasks. In order to improve their performance on it and reveal its law of cognitive processing, scholars have done many researches from the point of content effect, probabilistic representation and the reasoners’ characteristics, but the cognitive resource of working memory was rarely involved. This research was trying to study the relationship between working memory and Bayesian reasoning from the angle of the working memory span(WMS)and available resource of working memory with different frequency formats for further discussion.Two experiments were designed. In experiment 1, 72 of 120 college students were selected and divided into high WMS groupand low WMS group after finishing the operation span task, the experimental design was 2(WMS: high vs. low)×2(frequency formats: frequencies vs. probabilities). Each subject was asked to finish 5 Bayesian reasoning questions. In experiment2, we examined the relationship between central executive load and performance on Bayesian reasoning tasks in two kinds of question formats, and experimentally manipulated working memory resources in a dual task paradigm. The results were as followed:(1) Working memory was one of the most important influencing factors on Bayesian reasoning, generally speaking, the processing method of high WMS subjects was better than low WMS subjects, however, when the task was too difficult, the advantage of high WMS subjects may not exist. Central executive load had influence on Bayesian reasoning, specially, the magnitude of this effect was accompanied by an increase in working memory load intensity. When compared with the Bayesian performance in high load condition, the subjects performed better in low load condition and performed the best in no load condition.(2) The effect of Bayesian facilitation existed, but not found in all kinds of circumstances. We only found the effectof Bayesian facilitation on high WMS subjects and in low load condition and no load condition. These results consistently indicated that performance on classical Bayesian reasoning tasks depends on participants’ available working memory resources.(3)The findings lend credence to Nested Sets Theory and contradicted Ecological Rationality Framework.