Guidance Methods For Individual-group Interaction Behaviors Hindering Group Decision Making

The rapid development of society and technology has led to the segmentation of the fields.This trend causes the fact that the individual’s knowledge is limited.The limited information makes it difficult to ensure the reliability and accuracy of decision results,especially for complex decision problems.The group effectively improves this limitation,and the group is wiser than the individual.The reason is that each individual with different opinions can add information about the decision problem,enabling the group to analyze the decision problem comprehensively.Because of its significant advantages over individuals,the wisdom of crowds is valued in many fields,such as healthcare,public administration and e-commerce,and helps organizations and companies improve performance.However,in the actual decision-making process,it is difficult for individuals to maintain their independence.Individuals are easily influenced by other individuals and interact with the group.The situation violates the precondition for the wisdom of crowds,which negatively affects group decision results.First,there is the individual-group global interaction behavior that negatively affects group decision results.When there is disagreement among individuals in the group,the interaction between the individuals and the group facilitates intra-group coordination and the achievement of satisfactory decision results.Such global interactions between individuals and the group occur during the consensus reaching process.When the group forms a consensus on a wrong opinion,the group opinion has a systematic bias that negatively affects group decision results.Second,there is the individual-group local interaction behavior that negatively affects the group decision results.Individuals with bounded confidence are influenced by individuals with similar opinions,forming opinion evolution.Opinion evolution causes two negative effects on group opinion,which are disagreement and polarization of group opinion and bias of group opinion.Divergence and polarization lead to conflicts among individuals and generate invalid and unreasonable group opinions.And the bias of group opinion can cause group decision results deviate from the actual situation.Finally,the individualgroup global and local interaction behaviors that negatively affect group decision results can co-exist in group decision-making.Ranking is widely used as a method to address information overload,especially on digital platforms such as e-commerce platforms where ranking can significantly influence individual decisions.Since the nonpersonalized ranking methods use the historical purchase decisions of all individuals and personalized ranking methods exploit the historical purchase decisions of similar individuals,individuals interact globally and locally with the group under the influence of ranking methods.Under some conditions,the individual-group interactions under the influence of ranking methods can negatively affect the group decision results.Aiming at discussing the negative influence of the individual-group global interaction behavior,the individual-group local interaction behavior and the individualgroup compound interaction behavior composed of global and local interaction behaviors on group decision results,this dissertation deeply investigates the guidance methods of individual-group interaction behaviors to improve group decision results.The specific research of this dissertation includes the following four parts.1.The method of guiding the individual-group global interaction behavior in the consensus reaching process.Based on the optimization theory and the idea of supervised learning,a two-stage consensus reaching method is constructed to guide the individual-group global interaction behavior in the consensus reaching process.The method consists of the learning process of decision loss reference value based on the minimum decision error optimization model and the consensus reaching process based on the minimum decision loss difference consensus model.In this dissertation,we introduce this method to three-way group decision-making to effectively improve the accuracy of group decision-making.2.The method of guiding the individual-group local interaction behavior under opinion evolution with bounded confidence.Using the Markovian property of opinion evolution,we design an opinion management framework based on deep reinforcement learning to guide the individual-group local interaction behavior under opinion evolution.The framework includes a minimum adjustment cost consensus reaching method based on deep reinforcement learning and a minimum adjustment cost opinion guiding method based on deep reinforcement learning,which solves the divergence and the bias of group opinions formed by opinion evolution and effectively improves the quality of group opinions.3.The method of guiding the individual-group compound interaction behavior under the influence of ranking methods.An agent-based model and a discrete choice model are introduced to construct a simulation model that fits the actual situation of ecommerce platforms.The model connects the individual user’s purchase decision and the decision results of the user group.The simulation model is used to systematically analyze the influence of individual-group interactions on group decision results under different conditions and ranking methods.Then,the decision tree is trained based on the simulation results to help the platform managers to choose the appropriate ranking method according to the platform situation,which can guide individual-group interaction behaviors and improve the decision results of users.4.Research on the application of guidance methods of individual-group compound interaction behavior in e-commerce platforms.Since the susceptibility to ranking of users is different on different platforms,it is essential to study how to simulate users’ purchase decisions based on the actual situation of the platform,and help the platform choose the ranking method to guide the individual-group interaction behavior,which can improve group decision results.Therefore,based on the platform’s user purchase decision data,we discuss how the platform can estimate the simulation model at the individual level.Then,we implement the simulation analysis and provide suggestions for the platform to choose the ranking method to guide the individual-group interaction behavior in accordance with the platform’s situation.In summary,this dissertation focuses on the negative influence of individual-group interaction behaviors on group decision results and designs methods for guiding individual-group interaction behaviors.This dissertation not only enriches the theoretical research on promoting the wisdom of crowds,but also provides practical methods for managers to guide individual-group interaction behaviors and effectively improve group decision results.