Multi-decision-maker collaborative decision-making method, device and application

Abstract

The invention discloses a multi-decision-maker collaborative decision-making method, a multi-decision-maker collaborative decision-making device and application, relates to the technical field of man-machine collaboration, and is used for improving the reliability of multi-decision-maker collaborative decision-making. According to the invention, the adaptive weight of each decision maker in the decision task is determined based on the capability vector of each decision maker and the information to be decided. And after each decision maker scores each decision alternative item, weighting is carried out by using the decision weight of each decision maker to obtain a final score of each decision alternative item, and then a final decision alternative item is selected according to the final score. According to the method, the capability vector and the weight generation model parameters are subjected to end-to-end update learning according to the accuracy of the final decision, so that the reliability of collaborative decision is improved.

Description

Technical Field

[0001] The present invention relates to the field of human-machine collaborative technology, and in particular to a multi-decision maker collaborative decision-making method, a multi-decision maker collaborative decision-making device, and applications in image classification and management decision-making. Background Art

[0002] In a narrow sense, human-machine collaboration refers to collaborative decision-making by humans and artificial intelligence models on the same information to be decided, comprehensively considering the decision results of humans and artificial intelligence models, and leveraging human creative thinking and experience accumulation, as well as the high-speed computing power of artificial intelligence models, to improve the efficiency and quality of decision-making.

[0003] On this basis, the broad concept of human-machine collaboration can be derived, which includes not only the collaboration between humans and artificial intelligence models, but also scenarios in which all collaboration is done by humans and scenarios in which all collaboration is done by artificial intelligence models.

[0004] Existing human-machine collaborative decision-making can be roughly divided into two categories. The first category is to select the decision maker first, to decide whether the decision is made by humans or by artificial intelligence models, and the selected decision maker obtains the final decision result. This method completely distinguishes humans from artificial intelligence models, and does not comprehensively consider the decision advantages of humans and artificial intelligence models in the same dimension. The second category is that humans and artificial intelligence models first obtain alternative decision results, and then vote on the alternative decision results of the two to determine the final decision result. Although this method obtains the alternative decision results of humans and artificial intelligence models at the same time in the intermediate link, the final decision result is still one of the alternative decision results, and the decision result still does not integrate the decision advantages of humans and artificial intelligence models. Even if the alternative decision results of humans and artificial intelligence models are considered to be integrated, the decision-making ability of each decision maker on different decision-making information is not taken into account. It is assumed that the decision-making ability of each decision maker on different professions and fields is constant, which is obviously inconsistent with objective facts and will make the final integrated decision result unreliable. Summary of the invention

[0005] The object of the present invention is to provide a method, device and application for collaborative decision-making among multiple decision makers to improve the reliability of collaborative decision-making among multiple decision makers in order to solve all or part of the above-mentioned problems.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A multi-decision maker collaborative decision-making method, comprising:

[0008] Vectorize the information to be decided to obtain a representation sequence;

[0009] The capability vectors of each decision maker are obtained respectively, and after being concatenated with the representation sequence, the capability vectors are input into a weight generation model to obtain the decision weights of each decision maker respectively;

[0010] Obtaining scores of decision makers for each decision option of the information to be decided;

[0011] For each decision option, the score is weighted by using the decision weight of each decision maker, and the final score of each decision option of the information to be decided is calculated.

[0012] To solve the above problems, the present invention also provides a multi-decision-maker collaborative decision-making device, including a processor and a storage medium, wherein the storage medium stores a computer program, and when the processor runs the computer program in the storage medium, it executes the above-mentioned multi-decision-maker collaborative decision-making method.

[0013] The present invention also provides a method for image classification using the above-mentioned multi-decision maker collaborative decision-making method, wherein the information to be decided is the classification of the target image, and the decision options are preset decision options.

[0014] The present invention also provides a method for making management decisions using the above-mentioned multi-decision maker collaborative decision-making method, wherein the information to be decided is to determine decision options for a target decision scenario.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0016] The present invention introduces energy vectors of different decision makers in collaborative decision-making. The energy vector indicates the decision-making ability of the decision maker to deal with decision information, thereby generating a decision weight equivalent to the decision-making ability of the decision maker through a weight generation module, and then weighting the decision results (i.e., scores) based on the decision weights of different decision makers, and further calculating the final score of each decision option, which not only takes advantage of multi-person decision-making, but also fully considers the decision-making ability of each decision maker, making the final decision result more objective and reliable. In addition, the energy vector of the decision maker introduced in the present invention can also be designed as a learnable energy vector. By learning the features of the decision information sample, the energy vector of each decision maker is optimized, and a more complete representation of decision-making ability is learned, which further improves the reliability of the final decision result. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0018] Figure 1 It is a flow chart of the multi-decision maker collaborative decision-making method provided in an embodiment of the present application.

[0019] Figure 2It is a flowchart of dimensional alignment of input and output vectors of the weight generation model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0021] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other equivalent or alternative features having similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0022] In response to the problem that the existing human-machine collaborative technology considers the decision results of humans and artificial intelligence models independently, or integrates them through a stereotyped fixed fusion method, making the collaborative decision results unreliable, the embodiments of the present application provide a multi-decision maker collaborative decision-making method, device and application, which optimizes the contribution of different decision makers to the collaborative decision results by introducing capability vectors indicating the decision-making capabilities of different decision makers, thereby improving the reliability of collaborative decision-making.

[0023] In some embodiments, Figure 1 The multi-decision maker collaborative decision-making method includes the following process:

[0024] S1. Vectorize the information to be decided to obtain a representation sequence.

[0025] The information to be decided describes the decision background and the corresponding decision problem. In different scenarios, the information to be decided has different forms.

[0026] For example, in a problem decision scenario, the information to be decided is usually text information. For this type of information to be decided, the decision background and decision problem described in the information to be decided are segmented, and each segmented word is converted into an embedding vector through word embedding. The embedding vectors are sequentially connected to obtain a representation sequence of the information to be decided.

[0027] In some classification decision scenarios, such as image recognition scenarios, the information to be decided contains not only text information but also image information. For this type of information to be decided, the image as the decision background is encoded, the decision question text is segmented and embedded, and finally the image encoding sequence and word embedding vector are connected to obtain the representation sequence of the information to be decided.

[0028] S2. Obtain the capability vector of each decision maker respectively, concatenate it with the representation sequence, and input it into the weight generation model to obtain the decision weight of each decision maker respectively.

[0029] The decision maker's ability vector indicates the decision maker's decision-making ability, which is a multidimensional continuous value vector used to comprehensively measure the decision maker's ability related to the decision-making task. In the embodiment of the present application, whether the decision maker is a human or an artificial intelligence model, a unified dimension is used to measure the decision maker's decision-making ability, without considering humans and artificial intelligence models separately. The decision maker's energy vector determines the decision weight of its participation in the decision.

[0030] In some embodiments, the capability vectors of each decision maker are all preset capability vectors, or are all learnable capability vectors. The so-called preset energy vector, that is, the energy vector set in advance for each decision maker, does not change its value. For example, the energy evaluation index is used to construct the capability vector of the decision maker, and the energy vector is constructed by taking the professional knowledge question and answer scores, mathematics test scores, intelligence test scores, etc. as components. As an implementation method that is more favorable to the reliability of the final collaborative result, the use of a learnable energy vector will continuously optimize its value to better match the decision-making ability of the decision maker. For the learnable energy vector, it can be continuously optimized on the basis of a randomly initialized energy vector, or it can be continuously optimized on the basis of a certain set initial value, for example, it is continuously optimized on the basis of the energy vector constructed by the energy evaluation index in the previous text. The learnable energy vector can learn a more complete representation of decision-making ability, thereby obtaining a decision weight that is more matched with the decision-making ability of the decision maker, and further improving the reliability of collaborative decision-making.

[0031] As a feasible approach, the above-mentioned learnable ability vector is trained using decision information samples with score labels with the goal of minimizing the distance between the final score of the decision information sample and the corresponding score label.

[0032] The so-called decision information sample, that is, the historical decision information sample, has obtained the final decision result for its decision options. Based on the learning of a sufficient number of decision information samples, the value of the energy vector is continuously updated to make the decision-making ability reflected more accurate.

[0033] The present application embodiment provides a method for learning an energy vector:

[0034] Assuming that there are m decision makers involved in the decision-making, there are m capability vectors, which are represented as {c 1 ,…,c m}.

[0035] Let θ represent the parameters of the weight generation model, x represent the representation sequence of the information to be decided, and f(.) represent the operation of the weight generation model. The decision weights of each decision maker obtained by the weight generation model are expressed as:

[0036] w=(w 1 ,w 2,…,w m )=f(c 1 ,…,c m ,x,θ),

[0037] In the formula, w 1 ,w 2 ,…,w m They respectively represent the decision weights of the m decision makers.

[0038] In addition, in some preferred embodiments, in order to avoid the weight generation model from confusing the representation sequence and the energy vector, a segmentation embedding vector, usually an embedding vector of a text delimiter, may be added between the two embedding vectors.

[0039] Assuming that the number of decision options generated for the decision information sample is n, the scores of the n decision options given by m decision makers constitute an m×n dimensional matrix S∈R m×n The scoring matrix is ​​weighted by the decision weights w of m decision makers to obtain the final score s of n decision options. f ∈R 1×n for:

[0040] s f =wS.

[0041] To facilitate the operation of the training process, Softmax is used to normalize the final score:

[0042]

[0043] represents the final score of the i-th decision alternative.

[0044] The one-hot scoring labels of N decision information samples are represented as {y (1) ,…,y (i) ,…,y (N)},i∈[1,N], represents the kth position of the ith evaluation label. For N decision information samples, the corresponding scoring matrix {a (1) ,…,a (i) ,…,a (N)},i∈[1,N],

[0045] The goal is to train the energy vector to minimize and distance.

[0046] In some embodiments, the distance between the final score of the decision information sample and the corresponding score label is measured using the cross entropy loss. The cross entropy loss is expressed as:

[0047]

[0048] When there are enough decision information samples, the optimization of the cross entropy loss can update the values ​​of each ability vector.

[0049] The weight generation model in the embodiment of the present application adopts a model that maps a vector sequence to a few weight values. A variety of basic network structures can be used, such as encoder structures including BERT, Roberta, Longformer and other Transformers, decoder models similar to GPT, Palm, Transformer-XL and other Transformers, models similar to BART, XLNet and other models with both encoding and decoding capabilities, and traditional deep learning or machine learning model architectures with text processing capabilities such as RNN, LSTM, GRU, bi-LSTM, bi-GRU, etc.

[0050] In addition, when the length of the decision maker's capability vector is inconsistent with the length of the representation sequence of the information to be decided, it is impossible to directly concatenate the vectors. Figure 2 As shown in the figure, a first linear layer is configured before the weight generation module to unify the dimensions of each capability vector and the representation sequence before splicing. In addition, for the selected basic network structure, the dimension of the output layer needs to be configured so that each decision maker obtains a unique decision weight. The output dimension can be modified or a second linear layer can be configured after the output layer to obtain an output dimension that matches the number of decision makers.

[0051] In some embodiments, the parameter θ of the weight generation model can be updated with reference to the energy vector optimization method to further improve the accuracy of decision weight generation. That is, the weight generation model is trained by using decision information samples with scoring labels to minimize the distance between the final score of the decision information sample and the corresponding scoring label.

[0052] As a feasible approach, in the above embodiment of training the capability vector, the parameter θ of the weight generation model can be updated during the process of optimizing the cross entropy loss and updating the values ​​of each capability vector.

[0053] In a specific implementation, the value of the ability vector and the parameter θ of the weight generation model can be updated simultaneously or alternately during the cross entropy loss optimization process.

[0054] S3. Obtain the scores of each decision maker for each decision option of the decision information.

[0055] In some embodiments, each decision option is a decision option generated by each decision maker for the information to be decided, or is a preset decision option.

[0056] The preset decision options, that is, the decision options that have been set and determined for the information to be decided, are a priori settings, valid for any decision maker, and are not subject to the decision maker's subjective influence. The decision maker only needs to score these decision options separately. For example, in some image classification scenarios, the classification category of the target image is set in advance.

[0057] The decision options generated by the decision maker may be generated by the decision maker who participated in the decision, or by other decision makers who did not participate in the decision. The difference from the preset method is that the generated decision options are subject to the subjective influence of the decision maker. This is not limited in the embodiments of the present application.

[0058] In some embodiments, the decision options generated by the decision maker may include the following forms:

[0059] 1) Generated independently by each decision maker. For example, when both humans and AI models are decision makers, decision options are generated by humans and AI models separately and then aggregated.

[0060] 2) Humans generate examples of some decision options, and then the artificial intelligence model generates a series of other decision options based on the examples.

[0061] 3) The artificial intelligence model generates some decision options, and then humans refer to these decision options and generate more decision options based on their inspiration.

[0062] Each decision maker needs to score each decision option separately to reflect each decision maker's judgment on the degree to which each decision option conforms to the information to be decided.

[0063] S4. For each decision option, weight its score using the decision weight of each decision maker, and calculate the final score of each decision option of the information to be decided.

[0064] For each decision option, after obtaining the score of each decision maker for the decision option, the decision weight of each decision maker is used for weighted processing to calculate the final score of the decision option. As a feasible way, the score can be weighted summed or weighted averaged. Among them, when calculating the weighted average, the decision weight of each decision maker is the normalized decision weight.

[0065] S5. Select the final decision option based on the final score of each decision option.

[0066] After obtaining the final score of each decision option, each decision option can be sorted according to the final score. For a single-choice decision scenario, the decision option with the highest final score can be selected. For a multiple-choice decision scenario, a corresponding number of decision options with the highest final score can be selected. Of course, based on the final score of each decision option, other rules can also be used to determine the final decision option.

[0067] The embodiment of the present application also provides a multi-decision maker collaborative decision-making device, which includes a processor and a storage medium. The storage medium stores a computer program, and when the processor runs the computer program in the storage medium, it executes the multi-decision maker collaborative decision-making method of the above embodiment.

[0068] The embodiment of the present application also provides a method for image classification using the multi-decision maker collaborative decision-making method of the above embodiment. In this method, the information to be decided is the classification of the target image, that is, it includes the decision background of the target image and the decision problem of the classification. For this scenario, each decision option is a preset decision option, that is, each decision maker can directly score the preset decision option without generating new decision options. The preset decision options can be a determined set or a most likely subset of all possible options given by the decision maker.

[0069] For example, the decision problem may be the activity state of a person in an image (state A, state B, etc.), or which number in the image is 0 to 9.

[0070] In this scenario, the AI ​​model can use visual models such as ResNet and Vision-Transformer to determine the scores of different categories based on the probability distribution of different categories in the output layer. For human decision makers, they may directly identify one or a few categories. In this case, the scores can be determined based on the human decision maker's confidence in a few categories, while the scores of other categories are set to zero.

[0071] The present application example also verifies the effectiveness of the classification method:

[0072] The experiment uses a pre-trained wide resnet-18 as a vectorization model for decision information and an artificial intelligence model for scoring decision options. At the same time, the human decision-making process is simulated as follows: human experts have one or k expertise labels. When the true label of the input decision information sample is in the human expert's expertise label set, the human expert can accurately judge and give a prediction of the true label; when the true label of the input sample is not in the human expert's expertise label set, the human expert has a certain probability p of giving the correct label, and the remaining probability 1-p gives a randomly selected label. The decision weights corresponding to human experts and artificial intelligence models can be obtained through the weight generation model. When a total of m human experts and artificial intelligence models are used in collaboration, the initial ability vector is selected as an m+1-dimensional 0-1 one-hot vector, in which the i-th decision maker corresponds to a 0-1 vector with the i-th position being 1. By end-to-end learning of a post-linear layer, the 0-1 vector corresponding to a specific decision maker can give the ability vector of the corresponding decision maker after linear transformation. The label information in the CIFAR10 dataset is used as the training label. Figure 1 The model shown is trained end-to-end. For different values ​​of the number k of human expert ability expertise, the classification accuracy Test Accuracy of the trained collaborative model on the test set is shown in Table 1. The embodiment of the present application compares the method of the present application with several baseline models. These baseline models consider selecting a human decision maker or an artificial intelligence model to make the final decision instead of weighted summing the output probabilities. The results show that the method provided by the present application exceeds these baseline models at almost all values ​​of the number k of expertise.

[0073] Table 1 Accuracy comparison of this solution and the baseline model on the CIFAR10 dataset with different numbers of expertise

[0074] k 1 2 3 4 5 6 7 8 9 10 This application 95.53 95.85 96.16 97.21 97.66 98.47 98.96 99.25 99.66 100 L.5CE 94.63 95.26 95.53 95.89 96.31 98.08 98.71 99.14 99.42 100 Confidence 90.47 90.56 90.71 91.41 92.52 94.15 95.5 97.35 98.05 100 OracleReject 89.54 89.51 89.48 90.75 90.64 93.25 95.28 96.52 98.16 100 [MPZ18] 90.4 90.4 90.4 90.4 90.4 90.4 90.4 94.48 95.09 100

[0075] Considering different values ​​of the correct probability p under non-expert categories, the experiment is still conducted on the CIFAR10 dataset. The classification accuracy is shown in Table 2. It can be seen that the proposed method still has a stable advantage over the baseline model Confidence, and the baseline model's performance decreases when the comprehensive accuracy of human experts is close to that of the model. This reflects, to some extent, the shortcomings of choosing humans or artificial intelligence models to make decisions compared to using fusion weighted scoring.

[0076] Table 2 Comparison of the accuracy of this scheme and the baseline model on the CIFAR10 dataset under different correct probability values ​​of non-expert categories

[0077] 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 This application 97.66 97.79 97.94 98.09 98.35 98.32 98.6 98.78 98.75 99.14 100 Consistent[6] 96.31 96.58 96.9 97.15 97.13 97.29 95.23 93.88 92.94 95.6 100

[0078] Furthermore, consider that each human expert has only one different expertise category, that is, the expertise category of expert 1 is the first category of pictures with label 1, the expertise category of expert 2 is the second category of pictures with label 2, and so on. Since most traditional methods cannot use end-to-end deep learning models to handle the problem of multi-decision maker collaboration, the experiment compares the accuracy of the system with only different numbers of experts and different numbers of experts and artificial intelligence models working together, and the results are shown in Table 3. The results show that the method of the present application can effectively handle the collaborative decision-making of multi-expert systems, and the accuracy almost increases monotonically with the increase in the number of human experts. When the number of human experts is small, the accuracy is more obvious with the increase in the number of human experts.

[0079] Table 3 Accuracy of this scheme on CIFAR10 dataset with different number of experts (each expert has a different expertise category)

[0080] 0 1 2 3 4 5 6 7 8 9 10 AI+experts 93.5 95.8 96.76 97.21 97.2 97.3 97.2 97.3 97.2 97.2 97.3 experts 0 19 25.7 31.2 35.6 39.3 41.9 45.7 48.9 52.7 56.1

[0081] The embodiment of the present application also provides a method for making management decisions using the multi-decision maker collaborative decision-making method of the above embodiment, wherein the information to be decided is to determine the decision options for the target decision scenario, which is usually a text describing the decision background of the target decision scenario and the decision problem for determining the decision options. Each decision option is a decision option generated by each decision maker for the information to be decided, such as possible coping strategies in a certain dilemma, expansion strategies that can be adopted, etc. Or each decision option is a preset decision option, such as deciding whether to provide a loan, whether to recruit or dismiss an employee, whether to enter a certain market, etc.

[0082] As an optional approach, in order to deal with complex management decision-making problems, a generative large language model with reasoning ability, such as GPT-4, can be used as an artificial intelligence model to participate in decision-making. The generation of decision options can also be given by GPT-4.

[0083] Compared with the prior art, the solution of this application generally has the following advantages:

[0084] a. By learning the decision-making ability vectors of different decision makers, more accurate decision weight allocation can be obtained through adaptive methods in each specific decision-making task. In this application, the allocation of decision weights depends not only on the decision-making ability index of the decision maker, but also on the information to be decided (such as decision background knowledge). For different decision information samples, different abilities are required to accurately determine the decision results. For example, red-green color blindness and color weakness have different abilities in identifying different pictures. There are also certain differences in the accuracy of strategic experts, accountants, and human resources experts in judging different types of management decision problems.

[0085] Based on real decision information samples, a data-driven approach is used to learn the ability vectors of different decision makers and the mechanism of decision weight distribution, which can make the decision system achieve more accurate decision results. Existing technologies directly determine whether humans or machines make decisions based on decision background information, and their accuracy and generalization ability are weak.

[0086] b. Use a unified capability vector to represent the decision-related capabilities of different human decision makers and different artificial intelligence models. The capability vector learned from a large amount of labeled decision information sample data is a powerful indicator for measuring the capabilities of humans and artificial intelligence models across species. It can be reused and has a wide range of application scenarios.

[0087] c. This application can be well adapted to scenarios where multiple humans and multiple artificial intelligence models make mixed decisions, that is, the decision makers can all be humans or all be artificial intelligence models, or a combination of different numbers of humans and artificial intelligence models, and the number of decision makers can be two or more. The existing technology considers relatively limited situations, only considering multiple humans, multiple artificial intelligence models, or one person and one machine.

[0088] d. Existing human-machine collaborative decision-making systems mainly focus on discrimination-type decision problems, while the method of the present application can be effectively used in generation-type decision scenarios, such as management decisions and unstructured decision scenarios where the decision options are a piece of text or a process.

[0089] e. After obtaining the capability vectors of multiple decision makers, this application can optimize the combination of the decision makers' choices.

[0090] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A collaborative decision-making method for multiple decision makers, characterized in that: include: Vectorize the information to be decided to obtain a representation sequence; The capability vectors of each decision maker are obtained respectively, and after being concatenated with the representation sequence, the capability vectors are input into a weight generation model to obtain the decision weights of each decision maker respectively; Obtaining scores of decision makers for each decision option of the information to be decided; For each decision option, the score is weighted by using the decision weight of each decision maker, and the final score of each decision option of the information to be decided is calculated.

2. The method for collaborative decision making by multiple decision makers according to claim 1, characterized in that: The decision options are decision options generated by the decision makers for the information to be decided, or are preset decision options.

3. The method for collaborative decision making by multiple decision makers according to claim 1, characterized in that: The capability vectors of each decision maker are all preset capability vectors, or are all learnable capability vectors.

4. The method for collaborative decision making by multiple decision makers according to claim 3, characterized in that: The learnable capability vector is obtained by training using decision information samples with scoring labels with the goal of minimizing the distance between the final score of the decision information sample and the corresponding scoring label.

5. The method for collaborative decision-making by multiple decision makers according to claim 4, characterized in that: The learnable capability vector is obtained by training with the preset energy vector as an initial value.

6. The method for collaborative decision making by multiple decision makers according to claim 1, characterized in that: The weight generation model is trained using decision information samples with scoring labels with the goal of minimizing the distance between the final score of the decision information sample and the corresponding scoring label.

7. The method for collaborative decision-making by multiple decision makers according to any one of claims 3 to 6, characterized in that: The cross entropy loss is used to measure the distance between the final score of the decision information sample and the corresponding score label.

8. A multi-decision maker collaborative decision-making device, comprising a processor and a storage medium, characterized in that: The storage medium stores a computer program, and when the processor runs the computer program in the storage medium, it executes the multi-decision maker collaborative decision-making method as described in any one of claims 1-7.

9. A method for image classification using the multi-decision maker collaborative decision-making method according to claim 1, wherein: The information to be decided is classification of the target image, and the decision options are preset decision options.

10. A method for making management decisions using the multi-decision maker collaborative decision-making method according to claim 2, wherein: The information to be decided is to determine the decision options for the target decision scenario.

Images