摘要
再认启发式利用再认线索进行决策。以往研究采用一致率、击中率、虚报率和区分指数来表示再认启发式使用,然而这些方法都存在局限。多项式加工树模型能够分离不同的认知加工过程,为了解决再认使用与知识使用的混淆,研究者提出一种多项式加工树模型r-model测量再认启发式的使用。本文将重点介绍r-model,具体包括r-model的内容、数据分析以及考虑个体差异的分层r-model。最后,从r-model的模型修正和边界条件两个方面提出未来研究方向。
Recognition heuristic(RH)is frequently used when human makes decisions.The RH,as one of the simplest heuristics,has attracted the attention of many researchers,which only uses recognition to make decisions.RH can be defined as follows:if only one of the two items is recognized,individuals infer that the recognized item has stronger evidence,therefore it should be chosen.For example,the individual is asked to judge whether City A or City B is larger.City A is a recognized city,and City B an unrecognized city.If the individual chooses the recognized City A as the larger city,it indicates that the individual uses the RH.Many measures have been used in RH.The measures include but are not limited to:adherence rate,discrimination index,hit rate and false alarm rate.These measures are limited in distinguishing the key concepts of recognition and knowledge.To overcome this limitation,we introduce measures from a parametric multinomial model in this paper.The model is called Multinomial Processing Tree(MPT)model.MPT model is a family of effective statistical model measures and analyzes both explicit and implicit cognitive processes.MPT modeling methods have been successfully applied in many subject areas such as cognitive psychology,cognitive neurology,game theory,sociology,artificial intelligence and artificial intelligence network.We have built a version of MPT model in this paper specifically for measuring relevant cognitive processes in RH.We name it r-model.The r-model introduced in this paper contains three sub-model with each of the models targeting at three different cases:two objects are recognized(knowledge case),only one object is recognized(recognition case),and none of the objects is recognized(guessing case).The premise of using R-model is to use the pair comparison task as an experimental paradigm.After introducing the model,we explain how would one analyze data using r-model.The strength of r-model is the capability of analyzing individual differences.Most existing methods can only analyze data at the group level,therefore ignoring individual differences.The degree RH used in human cognition is highly individualized.RH researchers start to pay more attention to individual differences.Two methods are commonly used when considering individual differences in RH studies:hierarchical Bayesian models and hierarchical latent-class approach.Our proposed r-model is MPT implementation of the above mentioned two methods.The method of hierarchical modeling is used to define a separate MPT model and individual parameter estimates for each subject.It assumes that the individual parameter estimates for these individual MPT models come from a common distribution.The hierarchical latent-class approach is to use a finite hybrid model and assumes that the subject belongs to a limited number of potential classes and the same class of subjects have the same parameters.At the end of the paper,we propose the future research direction in using r-model in RH research.Specifically,we propose(1)How to modify model parameters to consider other measures such as response times and heuristic strategies;(2)How would one consider non-cognitive factors such as environmental conditions and individual characteristics.
作者
匡子翌
王效广
彭顺
杨磊
胡祥恩
Kuang Ziyi;Wang Xiaoguang;Peng shun;Yang Lei;Hu Xiang'en(Key Laboratory of Adolescent Cyberpsychology and Behavior,Ministry of Education,and Schol of Psyehology,Central China Normal University Wuhan,430079;College of Humanities,Jiangxi University of Traditional Chinese Medicine,Nanchang,330004;Department of Psychology,The University of Memphis,Tennessee,38152)
出处
《心理科学》
CSSCI
CSCD
北大核心
2021年第2期496-503,共8页
Journal of Psychological Science
基金
江西中医药大学2019年新引进博士博士启动基金项目(2019BSRW005)的资助。
关键词
再认启发式
流畅启发式
多项式加工树
贝叶斯分层模型
srecognition heuristic
fluency heuristic
multinomial processing tree model
Bayesian hierarchical model