Research On Three-way Behavior Decision Making For Human-machine Collaboration

The human-machine collaboration is one of the core issues in the field of the intelligentdecision-making research.Combining the human thinking and machine intelligence,the reasonablehuman-machine collaboration can produce stronger hybrid intelligent systems and better executionresults,which has attracted wide attentions from domestic and foreign scholars.In the current fieldof human-machine collaboration,how to make full use of the potential value in imprecise anduncertain information to provide objective and scientific decision supports for human-machinecollaboration issues such as conflict analysis,task allocation,and channel matching has become thekey research topic that needs to be solved urgently.Three-way decision,as an important researchmethod of granular computing and knowledge discovery,can consider the uncertainty and risk inthe decision-making process and divide the decision-making domain through probability thresholds,such that it can provide a solid support for the uncertainty and complexity of decision-makingproblems in human-machine collaboration.However,there are few existing related theories andmethods of human-machine collaboration and three-way decision that combine psychology andbehavior science,which can characterize the mechanism of human-machine collaboration and thedecision-making process of three-way decision model reasonably.To this end,this paper combinesbehavior decision-making,three-way decision and fuzzy decision-making theories to conduct theresearch and exploration on the issues of conflict analysis,task allocation and channel matching inhuman-machine collaboration.The subjective risk preference description of the decision-maker isincorporated into the three-way decision model to better characterize the cooperation mechanismbetween human and machine.The specific research contents and results are summarized as follows:(1)Propose a prospect theory-based three-way decision model and a prospect theory-based three-way conflict analysis method.In the existing theories and researches on three-way decision,how to reflect and describe the risk attitude and preference of decision-makers in the decision-making process is a very critical issue.In the three-way decision model with decision-theoretic rough sets,the decision rules are constructed according to the principle of risk minimization,without considering the real risk attitudes and preferences of different decision-makers.To solve this problem,this paper proposes a prospect theory-based three-way decision model.The maximum-prospect-value decision rules are derived by calculating value functions and weight functions.Then on this basis,this paper analyzes and proves the existence and uniqueness of the model's thresholds.The analytical solution forms of the thresholds in two specific situations are calculated and the decision rules are further simplified according to the values of thresholds.In addition,the prospect theory-based three-way decision model is deduced to the problem of conflict analysis in human-machine collaboration.The prospect theory-based three-way conflict analysis method is constructed,and the corresponding decision rules are derived.Finally,the prospect theory-based three-way conflict analysis method is applied to a human-machine collaboration conflict analysis example,and the effectiveness and performance of the method are evaluated based on several experiments.(2)Propose a three-way decision model based on cumulative prospect theory and a TOPSIS(Technique for Order Preference by Similarity to an Ideal Solution)method based on the multi-source heterogeneous information table.Based on the prospect theory-based three-way decision model,this paper extends the weight function of the model to the form of the cumulative weight function.By comparing and summarizing the value relationship of all value functions,the mathematical expression of the cumulative weight function of the model is obtained.The general form of the maximum-prospect-value decision rules of the proposed three-way decision model is derived according to the value function and the cumulative weight function.On this basis,the existence and uniqueness of thresholds are analyzed and proved,and the decision rules are simplified based on the numerical solutions of thresholds.Furthermore,in the case that the information system does not have decision attributes,this paper constructs a TOPSIS method based on the multi-source heterogeneous information table to estimate the conditional probability of the three-way decision model based on cumulative prospect theory.By considering three data types,including real numbers,linguistic term variables and intuitionistic fuzzy numbers,the conditional probability of the model is evaluated by the proposed TOPSIS method.Finally,the three-way decision model based on cumulative prospect theory and the TOPSIS method based on the multi-source heterogeneous information table are applied to an example of human-machine collaboration task allocation.We illustrate the effectiveness and advantages of the three-way decision model based on cumulative prospect theory by comparing it with the three-way decision model with decision-theoretic rough sets.(3)Propose a three-way decision model and a grey relational analysis method based on third-generation prospect theory and Z-numbers.On the basis of the previous two versions of prospect theory,third-generation prospect theory introduces uncertain reference points to better describe the risk attitude and preference of decision-makers.Compared with general fuzzy numbers,Z-numbers can better describe the reliability of data in the information system.Therefore,this paper further proposes a three-way decision model based on third-generation prospect theory and Z-numbers.In the model,all the data described by Z-numbers are converted into triangular fuzzy numbers,in order to expand the calculation of data more conveniently.With triangular fuzzy value functions and cumulative weight functions,the triangular fuzzy prospect value is calculated,and the maximum-prospect-value decision rules are established based on the corresponding ranking values.Considering that analytical solutions of the model's thresholds still cannot be directly calculated,this article analyzes the existence and uniqueness of the thresholds and simplifies the decision rules.Furthermore,according to the evaluation information in the Z-information system,this paper further proposes a grey relational analysis method based on third-generation prospect theory and Z-numbers to estimate the conditional probability of the three-way decision model.Finally,the three-way decision model and grey relational analysis method based on third-generation prospect theory and Z-numbers are applied to a human-machine collaboration channel matching problem.The advantages of the three-way decision model and grey relational analysis method based on third-generation prospect theory and Z-numbers are expounded by comparing with the three-way decision model based on the two previous versions of prospect theory and other conditional probability estimation methods.(4)Propose a three-way decision model and a TOPSIS method based on regret theory and interval type-2 fuzzy sets.The proposed three-way decision model and TOPSIS method can better describe the risk attitude and preference of decision-makers,and further improve the ability of the three-way decision model to deal with inaccurate and fuzzy problems.Under interval type-2 fuzzy environment,the interval-type-2 fuzzy regret-based utility functions are calculated by utilizing the interval-type-2 fuzzy outcome matrix.On this basis,the maximum-utility decision rules are determined by calculating the interval type-2 fuzzy expected utility function,and the decision rules are further derived according to the ranking values of the interval type-2 fuzzy expected utility functions.Considering that the above-mentioned conditional probability estimation methods cannot be directly applied to the interval type-2 fuzzy environment,this paper further proposes the TOPSIS method based on regret theory and interval type-2 fuzzy sets and utilizes this method to consider different decision-makers' decision preferences.At the same time,through a summary of the key decision-making steps of the three-way decision model and TOPSIS method based on regret theory and interval type-2 fuzzy sets,the detailed steps and schematic diagram of the overall decision-making process of the model are given.Finally,the three-way decision model and TOPSIS method based on regret theory and interval type-2 fuzzy sets are applied to the example of the human-machine collaboration channel matching problem,and the accuracy of all decision results is analyzed and verified through the public data sets.In brief,this article combines behavior decision-making,three-way decision and fuzzy decision-making theories to explore related issues of human-machine collaboration.This not only enriches the research of behavior decision-making,three-way decision and fuzzy decision-making theories,but also broadens the scope of applications of these theories.