Font Size: a A A

Research On Dynamic Model Of Learning Behavior Based On Social Cognitive Theory

Posted on:2021-04-23Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z H YanFull Text:PDF
GTID:1487306350468584Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
The rapid development of information technology has changed the way people behave and bring an understanding of human learning behavior into a new era.Social Cognitive Theory(SCT)believes that changes in people's beliefs and behaviors result from the combined effects of observational learning,personal learning,and self-regulation.SCT is the most influential social psychology theory that reveals the laws of human thought and behavior.In recent years,with the rise of complex networks and complexity science research,SCT-based research is no longer limited to empirical research,and the dynamic model research on SCT has gradually emerged.Based on SCT's general perspective and appropriate model settings,these models investigated human behavior through the lens of dynamic systems.Many dynamic laws of human learning behavior that are not easy to find in empirical research are revealed by these model research,showing a tremendous guiding significance in predicting and interventing human learning behavior.However,while the SCT dynamic model has demonstrated its ability to explain human behavior dynamics to a certain extent,it also shows some problems and limitations.When describing human motivation mechanisms,linear dynamic system models based on SCT are hard to explain the non-linear characteristics of the human psychological process.Also,introducing social influence factors into the model from the micro-individual level is insufficient in interpreting the macro-rules of group behavior.Some homogeneous high-complexity individual models are difficult to generalize to complex networks,also reduces their ability to describe social group behavior.In response to the above problems and challenges,this thesis first conducts dynamic modeling research on individuals.Based on the proposed signal structure knowledge model and individual efficacy model,group belief dynamics and behavior dynamics are researched by combining complex networks with these individual models.People's belief in the state of the objective world determines how they interact with it.The work on human belief dynamics in this thesis includes:(a)Introducing Bayesian inference into individual belief updates,establishing the basic model of individual belief dynamics during the process of periodic observations of world signals by taking the posterior beliefs of the previous period as the prior beliefs of the current period.(b)Providing the conditions individual's signal structure knowledge should satisfy to achieve different learning results and classifying individuals and their signal structures into "negative," "conservative," and "radical" by conditions provided.(c)Examining conservative agents and radical agents in a non-Bayesian social learning model combines observational learning and direct learning,revealing different learning processes and results between conservative and radical network groups.(d)Conducting model simulation to show that conservatives in the network are the backbone of successful group learning,and the harmonic averaging belief aggregation rule has a better performance.Compared with other studies,this thesis's belief dynamics models are based on the assumption of imperfect signal structure knowledge and non-optimal belief aggregation rules.It shows a higher ability in describing the social network phenomenon such as the belief pinning effect,information cascading effect,and echo chamber.Many periodical human activities have a memory effect-this thesis research the dynamic model of periodical behavior from the individual to the group.The work on the behavior dynamics model includes:(a)Defining individual efficiency with the rewards rates at different levels of effort and establishing a mathematical model describing the individual efficacy.(b)Analyzing the behavior effect mechanism along two paths:individual efficacy and perceived efficacy,establishing an individual's self-efficacy belief-behavior performance model based on the efficacy model.From the model simulation,it is found that individuals can accurately learn their efficacy through repeated experience of activities.(c)Introducing the social interaction matrix to extend the individual's self-efficacy belief-behavior performance model to the network group and revealing the role of high efficacy agent in the preferential attachment networks simulation.(d)In naive collectivity,providing a way to measure collective efficacy by the distributed perception of collective efficacy.Compared with other studies,research in this thesis is based on a simple non-linear model of individual efficacy.The influence of social structural factors is investigated from the micro level to the macro level.With the help of complex network methods,the models show their advantages in describing group behavior mechanisms and measuring collective efficacy.This research is a fusion of social psychology and complex network theories and methods.Based on SCT's conceptual model,this thesis establishes and studies dynamic models describing changes in beliefs and behaviors.Along with the established dynamic models,conclusions covering both the micro and macro levels are also conducted.Some of the conclusions support SCT's general views,such as the leading role of high efficacy agents,the impact of rewards and punishments on behavioral performance,and the mechanism that false feedback affects self-efficacy beliefs.Some of the conclusions are model discoveries,such as the different influences of conservative agents and radical agents on the dynamics of social beliefs,better performance of harmonic averaging belief aggregation rule,and collective efficacy measurement by individual efficacy perception.Generally speaking,observational learning adjusts the network group beliefs to reach consensus,while direct learning determines the network group direction of behavior changing.Combined with the influence of network topology and agents of different roles,social networks show simple laws in complex dynamic processes.This thesis' research methods and conclusions provide a model basis for practice and later empirical research.Simultaneously,it can trigger people's deeper thinking about the beliefs and behavior dynamic of individuals and human society.
Keywords/Search Tags:Observation Learning, Bayesian Updating, Self-Efficacy, Dynamic Model, Complex Network
PDF Full Text Request
Related items