Human and AI collaborative large-scale group decision-making method and equipment based on trusted network

Abstract

The embodiment of the invention discloses a human and AI collaborative large-scale group decision-making method and equipment based on a trusted network. The method comprises the following steps: constructing a decision-making problem; performing initial evaluation on the alternative scheme through a human decision maker set and an AI decision maker set to obtain initial evaluation information; constructing a human and AI trusted network; establishing a continuously learned human and AI consensus reaching algorithm, and dynamically updating the human and AI trust network and the initial evaluation information through the human and AI consensus reaching algorithm to obtain final evaluation information; calculating the weight of each attribute in the alternative scheme and the weight of each decision maker according to the final evaluation information; the final score of each alternative scheme is calculated, and a final decision is generated; the decision-making response speed, precision and decision-making consistency can be effectively improved, and the method is suitable for the complex decision-making field.

Description

Technical Field

[0001] The present application relates to the technical field of complex decision-making in data processing, and specifically to a method and device for large-scale group decision-making based on trust networks by humans and AI collaboration. Background Art

[0002] Group decision-making (GDM) refers to the process in which multiple decision makers (DMs) collaborate to determine the optimal option from a set of alternative options. In the group decision-making process, after discussion, communication and collision of opinions among members, the final decision becomes a key link. It is not only related to the quality and feasibility of the decision, but also affects the group members' acceptance of the decision and the strength of its implementation.

[0003] In traditional group decision-making, the final decision is usually made individually by human decision-makers (HDMs). Due to the limitations of their knowledge level, computing ability, and personal emotions, human decision-makers often cannot find the optimal solution when making the final decision.

[0004] Incorporating data-driven Artificial Intelligence Decision-Makers (ADMs) into group decision-making can address the above issues to a certain extent. Among them, AI decision-makers equipped with large language models (LLMs) stand out in group decision-making due to their ability to interact with humans, powerful computing power, and high accuracy. However, AI decision-makers have problems such as insufficient explainability, limited training data, and ethical biases, which may lead to a lack of accuracy and credibility in their output results, which in turn causes humans to distrust and ignore the advice of AI.

[0005] Therefore, terms such as human-computer interaction, human-computer collaboration, human-computer interaction, and human-AI collaboration have emerged one after another, aiming to combine the respective advantages of HDMs and ADMs.

[0006] As one of the important research directions in the field of artificial intelligence, human-AI collaborative decision-making (Human-AI collaboration decision) technology has been widely used in modern equipment manufacturing fields such as intelligent manufacturing and unmanned driving, as well as in many industries such as medical diagnosis and financial investment.

[0007] Existing methods for collaborative decision-making between humans and AI can be divided into three main categories: 1. Human-led, AI-assisted decision-making model: In this model, human decision makers are responsible for making major decision plans, and AI (Artificial Intelligence) serves as an auxiliary tool to provide data analysis, pattern recognition, or optimization suggestions; for example, AI can help researchers draft scientific review articles or provide preliminary recommendations.

[0008] Its limitations are that AI plays a relatively passive role and cannot truly participate in the decision-making process; at the same time, humans may ignore the effective suggestions it provides due to insufficient understanding of AI.

[0009] 2. Decision-making model dominated by artificial intelligence and adjusted by humans: In this model, artificial intelligence takes on the main decision-making tasks, and humans intervene or adjust only when necessary; for example, in medical scenarios, AI can interpret medical images to provide diagnostic suggestions, which are then confirmed by doctors; in autonomous driving scenarios, AI controls in real time based on sensor input, and humans intervene in complex or emergency situations.

[0010] However, due to the low explainability of AI and the lack of understanding of complex social factors and ethical considerations, relying entirely on AI decision-making may bring potential risks, leading to wrong decisions or a crisis of social trust.

[0011] The above-mentioned research on human-machine collaborative decision-making is mainly conducted from the perspectives of "humans assisted by artificial intelligence" and "artificial intelligence assisted by humans", which may lead to the failure to maximize the capabilities of each; therefore, more and more scholars have begun to regard the two as equally important decision-makers and consider their two-way influence.

[0012] 3. Decision-making model of equal collaboration between humans and AI: In recent years, with the development of AI technology, some research has begun exploring models where humans and AI participate equally in decision-making. Specifically, humans and AI serve as joint decision-makers, influencing each other and making dynamic adjustments. For example, in smart healthcare, AI can discuss diagnostic plans with doctors; in intelligent manufacturing, humans and AI can collaborate to optimize production processes and adjust production strategies through continuous interaction.

[0013] The advantage of this approach is that it leverages the respective strengths of humans and AI, improving the comprehensiveness and rationality of decision-making. However, these studies still overlook the complexity of large-scale collaboration involving multiple people and multiple AIs in real-world environments. For example, the Internet of Vehicles (IoV) requires multiple AIs for perception, planning, and control, along with multiple human operators for oversight. Financial risk control and intelligent investment rely on multiple AI models for analysis, and multiple human experts for strategic guidance.

[0014] These scenarios highlight the increasing reliance on collaboration between multiple humans and multiple AIs, yet existing technologies lack the necessary mechanisms for such collaboration.

[0015] In the mechanism of collaboration between multiple humans and multiple AIs, existing research faces the following key issues: (1) There is no large-scale human-machine collaborative decision-making method involving multiple humans and multiple AIs. At the same time, there is limited research on human-machine consensus mechanisms that effectively integrate multiple human expertise and multiple AI knowledge. (2) The existing trust mechanism for human-machine collaborative decision-making is relatively weak. Some studies only focus on trust between HDMs or only on the trust of HDMs in ADMs. No research has systematically studied the trust mechanism for human-AI collaborative decision-making from multiple dimensions, including HDMs’ self-confidence, trust in other HDMs, and trust in ADMs. (3) The method of allocating decision makers' weights is still relatively simple. Currently, the main methods are subjective determination, objective determination, and a combination of subjective and objective determination. However, these methods often cannot explain the deviation from the real data, resulting in solutions that may be optimal in theory but ineffective in practice.

[0016] The above problems restrict the application and promotion of human-AI collaborative decision-making technology in complex environments; therefore, there is an urgent need to develop a more scientific and efficient large-scale human-machine group collaborative decision-making method and related equipment involving multiple people and multiple AIs, so as to improve the scientificity, efficiency and credibility of the multi-agent collaborative decision-making system. Summary of the Invention

[0017] In order to solve one of the above-mentioned technical defects, an embodiment of the present application provides a method and device for large-scale group decision-making based on trust network collaboration between humans and AI.

[0018] A first aspect of the embodiments of the present application provides a large-scale group decision-making method for collaboration between humans and AI based on a trust network, comprising the following steps: Constructing a decision problem; the decision problem includes: multiple alternatives, each of the alternatives corresponds to multiple attributes, and each of the attributes corresponds to corresponding weight information; Perform an initial evaluation of alternative options through a set of human decision makers and a set of AI decision makers to obtain initial evaluation information; Build a trust network between people and AI; Establish a continuously learning human-AI consensus-building algorithm, dynamically update the initial evaluation information through the human-AI consensus-building algorithm and the human-AI trust network, and obtain the final evaluation information; Based on the final evaluation information, calculate the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers; Based on the weight of each attribute in the alternative, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers, the final score of each alternative is calculated to generate the final decision.

[0019] In an optional embodiment of the present application, each decision maker in the set of human decision makers and each decision maker in the set of AI decision makers expresses initial evaluation information of alternative options through a dynamic probabilistic language term set.

[0020] In an optional embodiment of the present application, the expression of the set of human decision makers is: ; The expression of the AI ​​decision maker set is: ; Among the human decision makers, Individual decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; Among the AI ​​decision makers, AI decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; Among them, L represents the total number of human decision makers involved in decision-making; Q represents the total number of AI decision makers involved in decision-making.

[0021] In an optional embodiment of the present application, the building of a human-AI trust network includes: Construct a set of all decision makers based on the set of human decision makers and the set of AI decision makers; Constructing a set of trust edges between human decision makers , and the set of trust edges between human decision makers and AI decision makers ; Establishing a set of trust weights between human decision makers , and the set of trust weights between human decision makers and AI decision makers .

[0022] In an optional embodiment of the present application, the establishment of a continuously learning human-AI consensus-reaching algorithm dynamically updates the initial evaluation information through the human-AI consensus-reaching algorithm and the human-AI trust network to obtain the final evaluation information; including: Calculate the consensus level of all decision makers based on the initial evaluation information of all decision makers; Determine whether the consensus level of all decision makers meets the consensus threshold If satisfied, the initial evaluation information is output as the final evaluation information; if not satisfied, then: Divide all decision makers into multiple decision subgroups and calculate the consensus level of each decision subgroup; Determine whether the consensus level of each decision subgroup meets the consensus threshold ; If not, then: update the human and AI trust network and make feedback adjustments within the corresponding decision subgroup until the consensus threshold is met ; If it is satisfied, then: identify the key decision makers, update the evaluation information through the key decision makers, obtain the final evaluation information and output it.

[0023] In an optional embodiment of the present application, the process of dividing all decision makers into multiple decision subgroups includes:

[0024]

[0025] Among them, multiple decision subgroups are represented as: ; represents the qth decision subgroup.

[0026] In an optional embodiment of the present application, the step of calculating the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers based on the final evaluation information includes: Based on the improved inverse distance weighted method, the weight of each decision maker in the set of human decision makers and the weight of each decision maker in the set of AI decision makers are determined; Based on LDM, the subjective and objective weight indicators of the attributes in the alternatives are integrated to determine the comprehensive weight of each attribute in the alternatives.

[0027] In an optional embodiment of the present application, based on the LDM, the subjective and objective weight indicators of the attributes in the alternative solutions are integrated to determine the comprehensive weight of each attribute in the alternative solution, including: Calculate the subjective weight of the attribute according to the subjective calculation rules of the attribute weight based on the G1 method; Calculate the objective weight of the attribute according to the objective calculation rules of the attribute weight based on the CRITIC method; Based on the LDM-based attribute weight comprehensive determination method, the subjective weight and objective weight of the attribute are comprehensively calculated to obtain the comprehensive weight of the attribute.

[0028] In an optional embodiment of the present application, the objective weight of the attribute is calculated according to the objective calculation rule of the attribute weight based on the CRITIC method, including: Standardize the final evaluation information of the human decision maker set and the AI ​​decision maker set to obtain standardized data; The standard deviation of the normalized data is defined as the attribute variability; Calculate the correlation matrix used to evaluate the linear relationship between attributes based on the standardized data; Calculate the average correlation between each attribute and all other attributes based on the correlation matrix; The information content of each attribute is calculated based on the attribute variability and the average correlation representing the attribute conflict; According to the information content of the attribute, the objective weight of each attribute is calculated.

[0029] According to a second aspect of an embodiment of the present application, a computer device is provided, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0030] The trust network-based large-scale group decision-making method and device for human-AI collaboration provided in the embodiments of the present application can combine the experience of human decision makers with the computing power of AI decision makers, give full play to the advantages of both, and dynamically update the human-AI trust network and initial evaluation information through a consensus-reaching algorithm, adjust the trust relationship and decision weights, make group decision-making more flexible and adaptable, thereby improving the accuracy and rationality of group decision-making, and can effectively improve the response speed, accuracy and consistency of decision-making, thereby providing more powerful support for complex decision-making scenarios.

[0031] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the contents pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flowchart of a large-scale group decision-making method for collaboration between humans and AI based on a trust network provided in one embodiment of the present application; Figure 2 This is a flowchart of step S30 in a large-scale group decision-making method based on a trust network and collaborative AI provided by one embodiment of the present application; Figure 3 This is a flowchart of step S40 in a large-scale group decision-making method based on a trust network and collaborative AI provided by one embodiment of the present application; Figure 4 This is a flowchart of step S50 in a large-scale group decision-making method based on a trust network and collaborative AI provided by one embodiment of the present application; Figure 5 A flowchart of a large-scale group decision-making method based on trust network collaboration between humans and AI in a specific case of this application; Figure 6 This is a schematic diagram of the division results of decision-making subgroups in a specific case of this application; Figure 7 It is a decision-making subgroup in a specific case of this application Feedback and adjustment diagram; Figure 8 It is a decision-making subgroup in a specific case of this application Feedback and adjustment diagram; Figure 9 It is a decision-making subgroup in a specific case of this application Feedback and adjustment diagram; Figure 10 This is a schematic diagram of the changes in the trust network between humans and AI of all decision makers in a specific case of this application; Figure 11 It is a network diagram of key decision makers in a specific case of this application; Figure 12 It is a schematic diagram of the weights of key decision makers in a specific case of this application; Figure 13 This is a qualitative analysis diagram of the decision-making method in a specific case of this application and other decision-making methods. DETAILED DESCRIPTION

[0033] The solutions in the embodiments of the present application can be implemented using various computer languages, for example, the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0034] In order to make the technical solutions and advantages of the embodiments of the present application more clearly understood, the exemplary embodiments of the present application are further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, and are not an exhaustive list of all the embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other unless they conflict.

[0035] In the following explanation, in order to ensure the accuracy of citations and the fluency of reading, the key technical terms, abbreviations or abbreviations involved in the article are summarized and explained as follows: AI, Artificial Intelligence; HDM, Human Decision-Maker HDMs, Human Decision-Makers, all human decision makers; ADM, Artificial Intelligence Decision-Maker, AI decision maker; ADMs, Artificial Intelligence Decision-Makers, all AI decision makers; LLMs, Large Language Models, large language models; Human-AI collaboration Decision, human and AI collaborative decision-making; LDM, Level Difference Maximization, level difference maximization method; DPLTS, Dynamic Probabilistic Language Term Set, dynamic probabilistic language term set; IIDW, Improved inverse distance weighting, improved inverse distance weighting method; G1 method, or ordinal relation analysis method, is a subjective weighting method improved from the analytic hierarchy process (AHP), and is often used to determine the weights of each evaluation criterion in multi-criteria decision analysis. The CRITIC method, Criteria importance Through Intercrieria Correlation, determines the importance of criteria through the correlation between criteria. It is an objective weighting method used to determine the weights of evaluation indicators.

[0036] Some problems with existing technologies: like Figure 1 As shown, in the first aspect, the embodiments of the present application provide a large-scale group decision-making method for collaboration between humans and AI based on a trust network, including the following steps: S10, constructing a decision problem; the decision problem includes: multiple alternatives, each of the alternatives corresponds to multiple attributes, and each of the attributes corresponds to corresponding weight information; S20, performing an initial evaluation of the alternatives by the set of human decision makers and the set of AI decision makers to obtain initial evaluation information; S30, building a trust network between humans and AI; S40: Establish a continuously learning human-AI consensus-building algorithm, dynamically update the initial evaluation information through the human-AI consensus-building algorithm and the human-AI trust network, and obtain the final evaluation information; S50, calculating the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers based on the final evaluation information; S60, calculate the final score of each alternative plan based on the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers, and generate a final decision.

[0037] The trust network-based large-scale group decision-making method for human-AI collaboration provided in the embodiments of the present application can combine the experience of human decision makers with the computing power of AI decision makers, give full play to the advantages of both, and dynamically update the human-AI trust network and initial evaluation information through a consensus-reaching algorithm, adjust the trust relationship and decision weights, make group decision-making more flexible and adaptable, thereby improving the accuracy and rationality of group decision-making, and can effectively improve the response speed, accuracy and consistency of decision-making, thereby providing more powerful support for complex decision-making scenarios.

[0038] In some possible implementations of the first aspect, in S20, the expression of the set of human decision makers is: , the corresponding weight of each decision maker in the set of human decision makers is: ;in, Indicates the The corresponding weight of the individual decision maker is ; The AI ​​decision maker set The expression is: , the weight corresponding to each decision maker in the AI ​​decision maker set is: ;in, Indicates the AI decision makers, whose weights are .

[0039] Among the human decision makers, Individual decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; Among the AI ​​decision makers, AI decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; L represents the total number of human decision makers involved in decision-making, and Q represents the total number of AI decision makers involved in decision-making.

[0040] In certain possible implementations of the first aspect, each decision maker in the set of human decision makers and each decision maker in the set of AI decision makers expresses initial evaluation information of alternative options through a dynamic probabilistic language term set.

[0041] Specifically, the set of human decision makers obtains initial evaluation information of the alternative options by calculating the following formula: (1); Where S is the language term set, It's time The probabilistic language time variable on Time language terminology The probability of , yes the number of different language terms included; (2); in , . Each PLTV (probabilistic linguistic time variable) in should exhaust all the language terms in LTS; (3); Time series Indicates that the weight corresponding to each time point is ; The time series weight is calculated by formula (4); (4); Furthermore, the AI ​​decision maker set can obtain evaluation information of ADMs by talking to an open source large language model (LLM); for example, the LLM is queried using predefined prompts and enabled web search capabilities; the prompts used are as follows: "If you are an expert in a field, you have several options. , their basic information will be provided to you in the form of text, tables, pictures, etc., and you will be asked to evaluate them”; The evaluation rules are as follows: Please use " ”,…, “ ”, “ ”, “ ”,…, “ ", and give the probability of each evaluation; ; The dynamic probabilistic language term set derived from ADM is expressed as: (5); The subsequent calculation process is consistent with the process by which a set of human decision makers obtains initial evaluation information on alternative options (Formula (2) to Formula (4)), and will not be traced here.

[0042] In the embodiment of the present application, the human decision Initial evaluation matrix for the nth attribute of the mth alternative The expression is: ; The AI ​​decision maker Initial evaluation matrix for the nth attribute of the mth alternative The expression is: .

[0043] In certain possible implementations of the first aspect, S30, building a human-AI trust network, includes: S301, constructing a set of all decision makers based on the set of human decision makers and the set of AI decision makers; S302, build a trust edge set between human decision makers , and the set of trust edges between human decision makers and AI decision makers ; in, , represents the trust relationship between HDMs and is used to describe the trust transfer between HDMs; Represents human decision makers With human decision makers The trust relationship between ; , represents the trust relationship between HDM and ADM, and is used to describe the trust transfer between HDM and ADM; Represents human decision makers With AI decision makers The trust relationship between ; S303, establish a trust weight set between human decision makers , and the set of trust weights between human decision makers and AI decision makers ; in, , represents the trust weight set between HDMs, Represents human decision makers To human decision makers Trustworthiness; , represents the trust relationship weight set from HDMs to ADMs, Represents the human decision maker To AI decision makers degree of trust.

[0044] In certain possible implementations of the first aspect, S40, establishing a continuously learning human-AI consensus-reaching algorithm, and dynamically updating initial evaluation information through the human-AI consensus-reaching algorithm and the human-AI trust network to obtain final evaluation information, includes: S401, calculating the consensus level of all decision makers based on the initial evaluation information of all decision makers; The formula for calculating the consensus level of all decision makers is: (6); Among them, the represents the level of consensus among all decision makers; Represents human decision makers Evaluation matrix; represents the distance between the evaluation information of the μth decision maker and the evaluation information of the vth decision maker, and here represents the distance between the evaluation information of any two decision makers; and: ; S402: Determine whether the consensus level of all decision makers meets the consensus threshold ; S403, if satisfied, output the initial evaluation information as final evaluation information; S404, if not satisfied, then: divide all decision makers into multiple decision subgroups, and calculate the consensus level of each decision subgroup; The process of dividing all decision makers into multiple decision subgroups includes: (7); (8); The decision subgroups are: ; represents the qth decision subgroup; The calculation formula for the consensus level of the qth decision subgroup is: (9); and ; in, Represents decision subgroup The number of decision makers in .

[0045] S405: Determine whether the consensus level of each decision subgroup meets the consensus threshold. ; S406, if not satisfied, then: update the human and AI trust network and make feedback adjustments within the corresponding decision subgroup until the consensus threshold is met ; S407, if satisfied, then: identify key decision makers, update the evaluation information through the key decision makers, obtain the final evaluation information and output it.

[0046] In some possible implementations of the first aspect, in S406, the human and AI trust network is updated, and feedback adjustments are made within the corresponding decision subgroups until the consensus threshold is met. ;include: S406-1: Build a global trust propagation model based on the decision subgroups. The specific process includes: Assume: Decision subgroup The corresponding nodes are: ;in, Representation and AI decision makers the set of all human decision makers who interact with each other; set up: For all Towards The set of propagation paths, the weight vector of the corresponding propagation path is expressed as ,and 、 ; Decision-making subgroup Human decision makers Trust in AI decision makers; human decision makers The confidence level is , The standardized value is , The standardized value is , The standardized value is ; S406-2: Based on the global trust propagation model, feedback is adjusted on the trust network between humans and AI in the decision-making subgroup, and the evaluation information and consensus level of the decision-making subgroup are updated; The calculation formula for feedback adjustment of the human and AI trust network of the decision-making subgroup includes: (10); (11); In formula (10) After propagation through the u-th trust path, the human decision maker For human decision makers The trust value established; Indicates the trust value between two adjacent nodes on this path.

[0047] The calculation formula for updating the evaluation information of the decision subgroup is: (12); in, ; In formula (12), Represents the human decision maker after updating the global trust propagation model assessment information, Represents human decision makers assessment information, Represents human decision makers assessment information, It means that after the global trust propagation model, human decision makers For human decision makers The trust value, express S406-3, judge the consensus level of the updated decision subgroup. If it meets , then execute step S407; otherwise, execute step S406-4; in, represents the threshold value of the consensus level within the decision subgroup, Represents the human decision maker after adjustment by the global trust propagation model evaluation matrix.

[0048] S406-4, AI decision makers continuously optimize their evaluation capabilities through machine learning methods, update the trust network between humans and AI, and adjust decision information until ; Specifically include: The comprehensive similarity is calculated based on the trust similarity of AI decision makers, and the improved Sigmoid function is used to update the human and AI trust network; the calculation formula is: (13); (14); (15); (16); in, Represents human decision makers With human decision makers The evaluation similarity of Represents human decision makers With human decision makers The distance of the evaluation information, Represents human decision makers assessment information, Represents human decision makers With human decision makers For AI decision makers The trust similarity of Represents human decision makers With human decision makers For AI decision makers The trust value distance, Represents human decision makers For AI decision makers Trust value.

[0049] Indicates the trust value, which is between 0 and 1; The coefficient that represents the steepness of the control function; represents the threshold parameter for adjusting the center of the function; make, A comprehensive similarity threshold is used to determine whether a trust relationship can be established between two human decision makers; when , human decision makers Advised to trust human decision makers , continuously decreasing Value until .

[0050] In some possible implementations of the first aspect, in S407, identifying key decision makers, updating evaluation information through the key decision makers, obtaining final evaluation information, and outputting the final evaluation information may include: S407-1, Identify key decision makers within the decision subgroup; including: S407-1-1, introduces the degree centrality index to measure the influence of each HDM, defined as: (17); in, Represents the human decision maker To human decision makers Trust value; S407-1-2, calculate the number of decision subgroups to which the human decision maker belongs. The calculation formula is: (18); in, represents an indicator function; if: human decision maker Belongs to subgroup ,but ,otherwise ; Define the cross-subgroup human decision makers as: (19); Represents human decision makers The number of decision subgroups is greater than 1.

[0051] S407-1-3, the most influential human decision makers, cross-subgroup human decision makers, and AI decision makers are considered key decision makers ; The set of key decision makers is defined as: (20); S407-2: The key decision maker has another conversation with the AI ​​decision maker with whom he interacts to update the assessment information; key decision makers The updated evaluation matrix is , the calculation formula of the updated evaluation matrix of other decision makers is: (twenty one); Repeat the above steps until , obtain the final evaluation information of the alternative plan and output it; Among them, Final Evaluation Matrix Indicates that, and ; The final evaluation information of the qth AI decision maker is expressed as the final evaluation matrix Indicates that, and .

[0052] In traditional decision-making methods, trust relationships are often viewed as a static and single factor. However, in complex group decision-making, trust relationships constantly change with information updates and decision feedback.

[0053] In the embodiments of the present application, an innovative human-AI trust network based on multi-dimensional trust relationships is proposed, which comprehensively considers the self-confidence of HDMs, the mutual trust relationship between HDMs, and the trust of HDMs in ADMs, so that the trust network can be dynamically adjusted during the group decision-making process; the present application ensures that the trust relationship between different human decision makers and AI decision makers can reflect their influence on the decision in real time through a dynamic trust propagation mechanism. This dynamic adjustment mechanism not only improves the coordination between decision makers, but also enables group decisions to be adaptively optimized based on feedback, which is particularly suitable for complex environments involving a large number of participants in the decision-making process; in addition, this innovative trust network mechanism helps to maintain decision consistency and fairness in changing decision-making scenarios, thereby improving the reliability of the overall decision.

[0054] In certain possible implementations of the first aspect, S50, calculating, based on the final evaluation information, the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers, includes: S501, determining the weight of each decision maker in the set of human decision makers and the weight of each decision maker in the set of AI decision makers based on an improved inverse distance weighted method; Specifically, the core idea of ​​the Improved Inverse Distance Weighting (IIDW) method is that the closer a known point is to an unknown point, the greater its influence on the estimation of the unknown point, and the weight decays as the distance increases.

[0055] The final evaluation matrices for HDMs and ADMs are and ; The weights are recorded as: and .

[0056] Assumption: The real data is represented as , then: The distance between HDMs and real data is defined as: (twenty two); The distance between ADMs and the true data is defined as: (twenty three); The weight of HDMs is calculated as: (twenty four); The weight of ADMs is calculated as: (25); in, is a small positive number used to avoid zero denominators and acts as a smoothing or regularization factor.

[0057] S502 , based on the LDM, integrate the subjective and objective weight indicators of the attributes in the alternative solutions to determine the comprehensive weight of each attribute in the alternative solutions.

[0058] In certain possible implementations of the first aspect, the step S502, based on the LDM, integrating subjective and objective weight indicators of attributes in the alternative solutions to determine the comprehensive weight of each attribute in the alternative solutions, includes: S502-1, calculating the subjective weight of the attribute according to the subjective calculation rules of the attribute weight based on the G1 method; specifically including: Assumption: The attributes in the alternatives are ranked in order of importance as follows: ; Assign a relative importance ratio between adjacent attributes ; Initialize the weight of the least important attribute to , the weights of other attributes are recursively calculated, and the formula is: (26); The weights of all attributes are normalized, and the formula is: (27); Among them, the relative importance ratio Assignment is done through subjective assignment rules.

[0059] S502-2, calculate the objective weight of the attribute according to the objective calculation rules of the attribute weight based on the CRITIC method; specifically, the following steps are performed: S502-2-1, standardize the final evaluation information of the human decision maker set and the AI ​​decision maker set to obtain standardized data; the calculation formula for standardization is: (28); in, Indicates the Alternative option No. The true value of an attribute; S502-2-2, defines the standard deviation of normalized data as attribute variability; the standard deviation is calculated as: (29); in, represents the average value of the nth attribute; S502-2-3, calculate the correlation matrix used to evaluate the linear relationship between attributes based on the standardized data, and the calculation formula is: (30); S502-2-4, calculate the average correlation between each attribute and all other attributes based on the correlation matrix; the calculation formula is: (31); S502-2-5, calculate the information content of each attribute based on the attribute variability and the average correlation representing attribute conflict; the calculation formula is: (32); S502-2-6, calculate the objective weight of each attribute based on the information content of the attribute; The calculation formula is: (33).

[0060] S502-3, based on the LDM attribute weight comprehensive determination method, the subjective weight of the attribute and the objective weight of the attribute are comprehensively calculated to obtain the comprehensive weight of the attribute; The LDM method (Level Difference Maximization) integrates subjective and objective weights to obtain a more interpretable weight determination. Its calculation formula is: (34); in, represents the subjective weight of the attribute, represents the objective weight of the attribute, Indicates the comprehensive weight of the attribute; In some possible implementations of the first aspect, the relative importance of attributes can be quantified by a series of scores, wherein the relative importance ratio is As shown in Table 1: Table 1

[0061] In an embodiment of the present application, the weights of HDMs and ADMs are allocated based on the distance between the evaluation matrix and the real data by improving the inverse distance weighting method (IIDW). Compared with the traditional weight allocation method, the IIDW method can more flexibly respond to changes in actual data. In this method, the decision maker with a smaller distance has a greater influence on the final decision, ensuring that the contribution of each decision maker can accurately reflect its degree of conformity with the real data; through this method, the accuracy of group decision-making is greatly improved, and decision-making bias caused by improper weight allocation is effectively avoided; at the same time, the method has strong scalability and can flexibly adjust the weight allocation strategy according to different scenarios and data changes. Therefore, this method can be effectively applied in a variety of complex decision-making environments, especially in fields such as finance and energy that require precise decision-making.

[0062] In certain possible implementations of the first aspect, S60, calculating a final score for each alternative plan based on the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers, and generating a final decision, includes: Summarize decision information, calculate the final scores of alternative options, and select the best option.

[0063] (35); in, Indicates alternatives The final score, Represents decision makers The weight of Is a decision maker Final evaluation information.

[0064] Traditional AI-based decision-making systems, especially those using deep learning and complex algorithms, often face the "black box" problem, where the decision-making process is difficult to explain and understand, leading to a decrease in trust in the decision results. To address this issue, this paper uses the inverse distance weighting (IDW) method and a human-AI trust network to effectively address the transparency issue of AI decision-making to a certain extent.

[0065] In the embodiments of the present application, the weights and trust allocation process of all decisions can be clearly displayed, so that the process of forming each decision has a high degree of explainability. Specifically, the weight allocation of decision makers is dynamically calculated based on their fit with real data and trust relationship, and this process is traceable. Decision transparency not only enhances the social trust in group decision-making, but also reduces uncertainty in the decision-making process, allowing each participant to understand the basis and influence of their decision, further enhancing the reliability of the system. Through this method, the group decision-making system can provide clear decision-making basis for all parties, effectively avoid the "black box" effect in the AI ​​decision-making process, and enhance the acceptability and application scope of group decision-making.

[0066] Below, through a specific case, the application's large-scale group decision-making method for collaborative human and AI based on trust network is further explained.

[0067] like Figure 5 As shown, in a specific case, including: 5 alternatives (5 energy companies), labeled: ; Each alternative includes three attributes: financial status, legal compliance, and business development, labeled: ; By using Assess the creditworthiness of five energy companies; 16 human decision makers, labeled: ; 4 AI decision makers (widely used large language models, LLMs), labeled: ; Each human decision maker interacts with one or more of the four AI decision makers; Data were collected over five time periods (2019-2023), and the time points are marked as: .

[0068] In a specific case, the collaborative group decision-making process can be divided into several stages, including: In the first stage, the evaluation information of HDM and ADM is obtained by using the dynamic probabilistic language term set through formulas (1) to (5); for example: ; ; The time series weight is calculated by formula (4); For example: time series weights: ; Process the evaluation information through formulas (2) to (4); For example: Assessment information: ; .

[0069] In the second stage, the consensus level of all decision makers is obtained through formula (6); ; In a specific case, the consensus threshold is 0.9 like Figure 6 As shown in Figure 2, when the consensus level of all decision makers does not meet the consensus threshold, the 20 decision makers are divided into 4 decision subgroups using formulas (7) to (8). The division results of the 4 decision subgroups are shown in Table 2: Table 2

[0070] The consensus level of the four decision-making subgroups is calculated using formula (9), as follows: , , , ; It can be seen that: , , , therefore, it is necessary to adjust the decision subgroup , decision-making subgroup , decision-making subgroup Evaluation information in.

[0071] Specifically, the decision-making subgroup The initial trust values ​​of are shown in Table 3.

[0072] Table 3

[0073] For decision subgroups , the initial and updated human and AI trust networks are as follows Figure 7 As shown in (a) and (b), the updated trust network is calculated by formulas (10) to (11), and the updated evaluation information is generated by formula (12). After the update, the decision subgroup The consensus level has increased to 0.9196.

[0074] Furthermore, the decision-making subgroup The initial consensus level is 0.8420, and Table 4 shows the decision subgroups Initial trust value, Figure 8 (a) and (b) show the decision subgroups Initial and updated human-AI trust networks, after adjustments, decision subgroups The consensus level has increased to 0.8890.

[0075] Afterwards, the human and AI trust network is updated and the decision information is adjusted; including: calculating the comprehensive similarity through formulas (13) to (15), where the parameters are set to , ; The comprehensive similarity is converted into a trust value through the Sigmoid-based function defined by formula (16), and the parameters are set as , ; The updated human-AI trust network is as follows Figure 8 As shown in (c), the adjusted decision subgroup The consensus level reached 0.9210.

[0076] Table 4

[0077] Furthermore, the decision-making subgroup The initial consensus level is 0.8249, and its initial trust value is shown in Table 5.

[0078] Table 5

[0079] Decision-making subgroup The human and AI trust network within Figure 10 As shown, Figure 10 (a) shows the initial human-AI trust network diagram. Figure 10 (b) in the figure shows the updated human-AI trust network diagram calculated using formulas (10) to (11). The updated evaluation information is calculated using formula (12). The decision subgroup The consistency level increased to 0.9258.

[0080] Furthermore, the decision-making subgroup The initial consensus level is 0.901, .

[0081] At this point, all four decision-making subgroups have reached a consensus, and the overall consensus level is calculated using formula (6), .

[0082] Furthermore, the key decision makers in the decision subgroup are identified by formulas (17) to (20), and the key decision makers are as follows: Figure 11 As shown; the evaluation information is updated using formula (21), and after calculation, the final consensus level is 0.9264.

[0083] In the third stage, the weight of each attribute in the alternative options and the weight of each decision maker in the set of human decision makers and the set of AI decision makers are calculated.

[0084] The decision maker's weight is calculated by formulas (22) to (25), and the obtained decision maker's weight is as follows: Figure 12 shown.

[0085] Integrate the subjective and objective weight indicators of the attributes in the alternatives to determine the comprehensive weight of the attributes; the subjective weight of the attributes is calculated using formulas (26) to (27), specifically: The objective weight of the attribute is calculated using formulas (28) to (33), specifically: The comprehensive weight of the attributes is calculated by the horizontal difference maximization (LDM) method of formula (34), specifically: .

[0086] The fourth stage is to generate the final decision.

[0087] The final ranking of the alternatives determined by formula (35) is , the result is consistent with the ranking of real data. In a specific case, the real data is the ranking on Qichacha.

[0088] In a specific case, decision information from multiple human decision makers and multiple AI decision makers was aggregated, improving the efficiency and credibility of decision-making.

[0089] Table 6 is a qualitative analysis table of the decision-making method provided by this application and other decision-making methods (human only, AI only, multiple humans and single AI, and single human and multiple AI). Figure 13 This is a qualitative analysis diagram of the decision-making method provided by this application and other decision-making methods (human only, AI only, multiple humans and single AI, and single human and multiple AI).

[0090] In Table 6, the evaluation indicators for qualitative comparison include: MAE refers to Mean Absolute Error; MAPE is the Mean Absolute Percentage Error; SAE is the sum of absolute errors. MSE is Mean Squared Error; RMSE is the root mean square error (Root Mean Square Error); R2 coefficient of determination (R-Square); Explained Variance refers to explained variance; SSE is the sum of squares due to error or residual sum of squares.

[0091] Table 6

[0092] In Table 6, Human Only (HO) refers to a method that relies solely on personal experience, intuition, or team discussion to make decisions. Its decision-making principle relies entirely on a group of human decision makers to reach a consensus through discussion or experience-based judgment. The advantages of this method are that human decision makers are good at dealing with ambiguity and subjectivity and can weigh ethical and social impacts. The disadvantages are that human decision makers are susceptible to personal biases (for example, confirmation bias, groupthink) and are less efficient. It is often used in scenarios involving value or ethical decision-making (for example, corporate management culture, education policy).

[0093] In Table 6, AI (AO) alone is an autonomous decision-making method based on algorithms and data generated entirely by artificial intelligence systems. The decision-making principle is that an independent AI system makes decisions autonomously based on algorithms and data. The advantages of this method are: it can process large amounts of data in real time (for example, high-frequency trading, autonomous driving) and is not affected by human emotions or biases (for example, resume screening). The disadvantages are: complex models make it difficult to explain decisions and may amplify biases in the data. It is mostly used in scenarios of data-intensive tasks (for example, logistics planning) and rapid response solutions (for example, network defense).

[0094] In Table 6, multi-human and single AI (MHSA) is a method in which a team collaborates with a single AI system to make decisions. The decision-making principle is that multiple human decision makers use a single AI-assisted tool, where the AI ​​provides unified advice to support the decision-making process. The advantages of this method are: AI provides data support, humans make the final judgment (such as image analysis in medical diagnosis), and only one AI system needs to be maintained. The disadvantages are: a single AI may not be able to cover all dimensions of complex problems, and humans may over-trust the advice of AI. It is mostly used in scenarios where small to medium-sized teams have limited resources for tasks that require rapid analysis and human supervision (for example, financial risk assessment).

[0095] In Table 6, Single Human with Multiple AI (SHMA) is a method in which a human assists decision-making with the help of multiple AI systems. The decision-making principle is that a single human decision-maker integrates different suggestions from multiple AI systems to reach a final result. The advantages of this method are that multiple AIs can cover different functions, and individuals can quickly compare different AI recommendations. The disadvantages are that individuals need to understand and integrate the outputs of multiple AIs, which may be beyond their capabilities, and the advice provided by different AIs may conflict (for example, there are differences in legal advice). It is efficient in professional tasks (for example, research, content creation) and multidimensional analysis (for example, portfolio management).

[0096] In the embodiments of this application, multi-human and multi-AI (HMMA) is a method in which multiple people and multiple AIs participate in decision-making. Its decision-making principle is to achieve collaborative optimization by integrating dynamic collaboration between human decision-makers and AI systems, combining human judgment with the analytical capabilities of AI. The advantages of this method are: humans provide experience, creativity, and ethics; AI provides data analysis and pattern recognition; multiple AIs ensure continuous learning and reduce errors; and diverse human input prevents bias. Disadvantages are: the decision-making process can be lengthy, especially in cases of disagreement; the boundaries between human and AI responsibilities are unclear, which may lead to ethical disputes; and it is often used in decision-making solutions for complex tasks that require a balance between efficiency and accuracy (such as climate change policy and military strategy).

[0097] In summary, the present application can reflect the changes of each decision maker in real time through the adjustment of trust relationships in a changing decision-making environment, further improving the accuracy and rationality of group decision-making. The decision-making method in the present application is particularly suitable for complex decision-making scenarios that require multi-party collaboration, such as emergency response, cross-industry collaboration, smart medical care and other fields. In these environments, the present invention can effectively improve the response speed, accuracy and consistency of decisions, thereby providing more powerful support for complex decision-making scenarios.

[0098] It should be understood that, although the various steps in the flowchart are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps may be performed in other orders. Moreover, at least a portion of the steps in the figure may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but may be performed at different times. The execution order of these sub-steps or stages is not necessarily to be performed in sequence, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0099] In one embodiment, a computer device is provided. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements the above-mentioned method for large-scale group decision-making of humans and AI collaboration based on a trust network. It includes: a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements any step in the above-mentioned method for large-scale group decision-making of humans and AI collaboration based on a trust network.

[0100] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, any step in the above-mentioned large-scale group decision-making method for collaboration between humans and AI based on a trust network can be implemented.

[0101] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0102] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0105] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0106] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A large-scale group decision-making method based on trust network collaboration between humans and AI, characterized by: The following steps are involved: Constructing a decision problem; the decision problem includes: multiple alternatives, each of the alternatives corresponds to multiple attributes, and each of the attributes corresponds to corresponding weight information; Perform an initial evaluation of alternative options through a set of human decision makers and a set of AI decision makers to obtain initial evaluation information; Build a trust network between humans and AI; Establish a continuously learning human-AI consensus-building algorithm, dynamically update the initial evaluation information through the human-AI consensus-building algorithm and the human-AI trust network, and obtain the final evaluation information; Based on the final evaluation information, calculate the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers; Based on the weight of each attribute in the alternative, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers, the final score of each alternative is calculated to generate the final decision.

2. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 2 is characterized in that: Each decision maker in the set of human decision makers and each decision maker in the set of AI decision makers expresses initial evaluation information on alternative options through a dynamic probabilistic language term set.

3. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 2 is characterized in that: The expression of the set of human decision makers is: ; The expression of the AI ​​decision maker set is: ; Among the human decision makers, Individual decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; Among the AI ​​decision makers, AI decision makers The initial evaluation information of the nth attribute of the mth alternative is given by the initial evaluation matrix express; Among them, L represents the total number of human decision makers involved in decision-making; Q represents the total number of AI decision makers involved in decision-making.

4. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 3 is characterized in that: The construction of a trust network between humans and AI includes: Construct a set of all decision makers based on the set of human decision makers and the set of AI decision makers; Constructing a set of trust edges between human decision makers , and the set of trust edges between human decision makers and AI decision makers ; Establishing a set of trust weights between human decision makers , and the set of trust weights between human decision makers and AI decision makers .

5. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 4 is characterized in that: The method of establishing a continuously learning human-AI consensus-building algorithm dynamically updates the initial evaluation information through the human-AI consensus-building algorithm and the human-AI trust network to obtain the final evaluation information; including: Calculate the consensus level of all decision makers based on the initial evaluation information of all decision makers; Determine whether the consensus level of all decision makers meets the consensus threshold If satisfied, the initial evaluation information is output as the final evaluation information; if not satisfied, then: Divide all decision makers into multiple decision subgroups and calculate the consensus level of each decision subgroup; Determine whether the consensus level of each decision subgroup meets the consensus threshold ; If not, then: update the human and AI trust network and make feedback adjustments within the corresponding decision subgroup until the consensus threshold is met ; If it is satisfied, then: identify the key decision makers, update the evaluation information through the key decision makers, obtain the final evaluation information and output it.

6. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 5 is characterized in that: The process of dividing all decision makers into multiple decision subgroups includes: ； ； Among them, multiple decision subgroups are represented as: ; represents the qth decision subgroup.

7. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 6 is characterized in that: The method of calculating the weight of each attribute in the alternative plan, the weight of each decision maker in the set of human decision makers, and the weight of each decision maker in the set of AI decision makers based on the final evaluation information includes: Based on the improved inverse distance weighted method, the weight of each decision maker in the set of human decision makers and the weight of each decision maker in the set of AI decision makers are determined; Based on LDM, the subjective and objective weight indicators of the attributes in the alternatives are integrated to determine the comprehensive weight of each attribute in the alternatives.

8. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 7 is characterized in that: Based on the LDM, the subjective and objective weight indicators of the attributes in the alternatives are integrated to determine the comprehensive weight of each attribute in the alternatives, including: Calculate the subjective weight of the attribute according to the subjective calculation rules of the attribute weight based on the G1 method; Calculate the objective weight of the attribute according to the objective calculation rules of the attribute weight based on the CRITIC method; Based on the LDM-based attribute weight comprehensive determination method, the subjective weight and objective weight of the attribute are comprehensively calculated to obtain the comprehensive weight of the attribute.

9. The method for large-scale group decision-making based on trust network collaboration between humans and AI according to claim 8, characterized in that: The objective weights of the attributes are calculated according to the objective calculation rules of the attribute weights based on the CRITIC method, including: Standardize the final evaluation information of the human decision maker set and the AI ​​decision maker set to obtain standardized data; The standard deviation of the normalized data is defined as the attribute variability; Calculate the correlation matrix used to evaluate the linear relationship between attributes based on the standardized data; Calculate the average correlation between each attribute and all other attributes based on the correlation matrix; The information content of each attribute is calculated based on the attribute variability and the average correlation representing the attribute conflict; According to the information content of the attribute, the objective weight of each attribute is calculated.

10. A computer device comprising: The method comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

Images