Consensus decision-making method, device, equipment and program product

Abstract

The embodiment of the invention provides a consensus decision-making method and device, equipment and a program product. The method comprises the following steps: sending an initial questionnaire to each expert according to input information of a user; the input information comprises mail addresses and questionnaire contents of all experts, performing consistency evaluation on the received opinions of all the experts, determining an updated questionnaire based on a large language model when the consistency is not achieved, sending the updated questionnaire to all the experts until the opinions of all the experts achieve the consistency, and obtaining a consensus conclusion, in the consensus decision making process, the questionnaire is automatically adjusted and updated through the large language model, the accuracy of the adjusted and updated questionnaire is improved, the needed time is shortened, and therefore the efficiency and accuracy of the whole research process are improved.

Description

Technical Field

[0001] This application relates to the field of financial technology or other related fields, and in particular to a consensus decision-making method, apparatus, device and program product. Background Technology

[0002] The traditional Delphi method is a classic expert consensus-building approach that is widely used in policy making, technology forecasting, and medical decision-making.

[0003] The core advantage of the Delphi method lies in its systematic collection and aggregation of expert opinions through multiple rounds of anonymous questionnaires, thereby making the final results more scientific and impartial. Because experts remain anonymous, the influence of authority and group pressure is effectively weakened, ensuring the independence of expert opinions. Furthermore, the feedback provided after each round of surveys allows experts to adjust their views accordingly, gradually leading to a convergence of opinions and improving the quality of decision-making.

[0004] However, a key problem with the traditional Delphi method is its high dependence on human intervention. The summarization, classification, and analysis of expert feedback often require manual work, which not only consumes a lot of time and resources but may also be affected by the user's own subjectivity, reducing the objectivity of the analysis. As the scale of data increases, the difficulty of manual processing also increases significantly, thus affecting the efficiency and accuracy of the entire research process. Summary of the Invention

[0005] This application provides a consensus decision-making method, apparatus, equipment, and program product to solve the problems of low efficiency and poor accuracy in reaching expert consensus.

[0006] Firstly, this application provides a consensus decision-making method, including:

[0007] Initial questionnaires are sent to various experts based on the user's input information; the input information includes the email addresses of each expert and the content of the questionnaire.

[0008] A consistency assessment was conducted on the opinions received from all experts.

[0009] When no consensus is reached, an updated questionnaire is determined based on the large language model and sent to each expert until the opinions of all experts reach a consensus and a consensus conclusion is reached.

[0010] Secondly, this application provides a consensus decision-making device, comprising:

[0011] The sending module is used to send the initial questionnaire to each expert based on the user's input information; the input information includes the email address of each expert and the content of the questionnaire;

[0012] The evaluation module is used to assess the consistency of the received expert opinions.

[0013] The processing module is used to determine an updated questionnaire based on a large language model when no consensus is reached, and to send the updated questionnaire to each expert until the opinions of all experts reach a consensus and a consensus conclusion is obtained.

[0014] Thirdly, this application provides an electronic device, including: at least one processor and a memory;

[0015] The memory stores computer-executed instructions;

[0016] The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any of the first aspects.

[0017] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method as described in any of the first aspects.

[0018] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any of the first aspects.

[0019] The consensus decision-making method, apparatus, equipment, and program products provided in this application include: sending an initial questionnaire to various experts based on user input information, including the email addresses and questionnaire content of each expert; performing a consistency assessment on the received expert opinions; when no consensus is reached, determining an updated questionnaire based on a large language model; sending the updated questionnaire to each expert; and continuing until the experts reach a consensus and a consensus conclusion is reached. During the consensus decision-making process, the updated questionnaire is automatically adjusted using a large language model, improving the accuracy of the adjusted questionnaire and reducing the required time, thereby improving the efficiency and accuracy of the entire research process. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0021] Figure 1 A flowchart illustrating a consensus decision-making method provided in an embodiment of the present invention;

[0022] Figure 2 A flowchart illustrating an expert opinion consistency assessment method provided in an embodiment of the present invention;

[0023] Figure 3A schematic diagram illustrating a complete consensus decision-making method provided in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of a consensus decision-making device 40 provided in an embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention.

[0026] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize or refuse.

[0029] It should be noted that the consensus decision-making method, apparatus, equipment, storage medium and program products provided in this application can be used in the field of fintech, or in any field other than fintech. This application does not limit the application field of the consensus decision-making method, apparatus, equipment, storage medium and program products.

[0030] Currently, the traditional Delphi method can achieve consensus-based decision-making by summarizing the opinions of multiple experts and adjusting the questionnaire, allowing experts to adjust their views and thus bringing their opinions closer together to reach a consensus. However, existing methods require human intervention; that is, the process of summarizing expert opinions and adjusting the questionnaire needs to be handled manually, which results in problems of low efficiency and accuracy.

[0031] The consensus decision-making method provided in this application aims to solve the above-mentioned technical problems in the prior art.

[0032] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0033] Figure 1 This is a flowchart illustrating a consensus decision-making method provided in an embodiment of the present invention. The method can be applied to an intelligent agent, specifically a large language model-driven agent (i.e., an intelligent proxy), to achieve consensus decision-making based on the large language model. The method includes steps S101 to S103:

[0034] Step S101: Send the initial questionnaire to each expert based on the user's input information; the input information includes the email address of each expert and the questionnaire content.

[0035] Users can set different input information for different research questions. Input information can include the email addresses of various experts and the content of the questionnaire. Different research questions will correspond to different expert groups, and the questionnaire content will also differ. Optionally, the input information can also include a list of experts.

[0036] Optionally, upon receiving user input, the input can be validated, such as validating the format of each expert's email address. If the format does not meet the requirements, the user is instructed to re-enter the information.

[0037] Optionally, the questionnaire content can be formatted into an email template to generate an initial questionnaire, which can then be sent to each expert via email. Alternatively, the email sending function can be invoked to send an email containing the initial questionnaire to each email address.

[0038] For example, if a user enters the email addresses of 10 experts and a questionnaire about "the development trend of new energy vehicles", 10 emails can be automatically generated and sent to each expert.

[0039] Step S102: Conduct a consistency assessment of the received expert opinions.

[0040] After sending emails to each expert, you can also receive their feedback, allowing you to conduct a consistency assessment of the received expert opinions. This consistency assessment determines whether to end the questionnaire update process and arrive at a consensus conclusion.

[0041] Optionally, after receiving feedback emails from various experts, the email content can be processed into a uniform text format to facilitate consistency assessment.

[0042] Optionally, one can quantify and determine whether there is consensus among the experts on the questionnaire content, that is, whether their opinions tend to be consistent or there are significant differences.

[0043] Optionally, a high degree of consistency indicates that the experts' opinions are very unified, the conclusion is highly credible, and can be regarded as a consensus conclusion; a low degree of consistency indicates that the experts' opinions are scattered, suggesting that the questions corresponding to the questionnaire content are relatively complex and uncertain, requiring further discussion and demonstration.

[0044] Step S103: When no consensus is reached, an updated questionnaire is determined based on the large language model and sent to each expert until the opinions of each expert reach a consensus and a consensus conclusion is obtained.

[0045] After conducting a consensus assessment, it can be determined whether the opinions of the experts have reached a consensus. If they have reached a consensus, a consensus conclusion can be obtained. If they have not reached a consensus, the large language model can be used to automatically determine and update the questionnaire and send it to each expert to obtain their opinions again. The consensus assessment can then be conducted again, and the above process can be repeated to obtain a consensus conclusion.

[0046] Optionally, the disagreements among expert opinions can be analyzed based on a large language model to generate an updated questionnaire, which can reduce human intervention errors and improve accuracy.

[0047] Optionally, multiple large language models can be set up for separate analysis, and the updated questionnaires determined by the multiple large language models can be merged to improve the accuracy of the output results.

[0048] In summary, intelligent agents can automatically generate, distribute, collect, and parse questionnaires. The method presented in this application enhances the automation level of the Delphi method while improving the scientific rigor, accuracy, and efficiency of consensus formation, realizing an intelligent expert consensus construction scheme based on a large language model. This application not only improves the efficiency of reaching expert consensus but also ensures a more transparent, accurate, and reliable process, providing strong support for various fields requiring expert consensus decision-making.

[0049] The consensus decision-making method provided in this application includes: sending an initial questionnaire to various experts based on user input information; the input information includes the email addresses of each expert and the questionnaire content; performing a consistency assessment on the received opinions of each expert; when no consensus is reached, determining an updated questionnaire based on a large language model; sending the updated questionnaire to each expert; and continuing until the opinions of each expert reach a consensus and a consensus conclusion is obtained. In the consensus decision-making process, the updated questionnaire is automatically adjusted through a large language model, which improves the accuracy of the adjusted updated questionnaire and reduces the required time, thereby improving the efficiency and accuracy of the entire research process.

[0050] Figure 2 This is a flowchart illustrating an expert opinion consistency assessment method provided in an embodiment of the present invention.

[0051] Optionally, a consensus assessment may be conducted on the received expert opinions, including:

[0052] Step S201: For any expert opinion, determine the vector corresponding to the expert opinion based on the large language model;

[0053] Step S202: Calculate the cosine similarity between any two vectors, and calculate the average of all cosine similarities;

[0054] Step S203: When the average value is greater than the preset value, it is determined that the opinions of the experts have reached a consensus; when the average value is less than the preset value, it is determined that the opinions of the experts have not reached a consensus.

[0055] When assessing the consistency of expert opinions, a method based on cosine similarity can be used. First, each expert opinion is converted into a high-dimensional numerical vector. Then, the cosine similarity between each pair of expert opinions is calculated, and the average of these similarities is taken. This average is then compared with a preset value to determine whether consistency has been achieved.

[0056] Specifically, each expert opinion can be output as a vector based on a large language model. This vector is a fixed-length, high-dimensional numerical vector, also known as an embedding vector. Vectors corresponding to semantically similar texts will have more similar directions in the vector space.

[0057] After determining the vectors corresponding to each expert opinion, cosine similarity can be calculated for each pair of vectors. Then, the average of all cosine similarities is calculated. When the average is greater than a preset value, it can be determined that the expert opinions are in agreement; when the average is less than the preset value, it can be determined that the expert opinions are not in agreement. The calculated average represents the overall index of the degree of consensus among the expert opinions. This calculation can detect the convergent or divergent parts of expert opinions.

[0058] Optionally, this application does not impose specific limitations on the setting of preset values, which can be set based on experience. In scenarios requiring strict consensus, the preset value should be higher; in scenarios requiring only general consensus, the preset value should be lower.

[0059] By calculating cosine similarity, the consistency of opinions among various experts can be accurately measured.

[0060] Optionally, the updated questionnaire can be determined based on a large language model, including:

[0061] Based on the expert opinions, multiple prompts and their corresponding weight values ​​are determined.

[0062] Logical flaws were identified based on the expert opinions mentioned above;

[0063] The multiple prompts, their corresponding weight values, and the logical flaws are input into the large language model to obtain the updated questionnaire.

[0064] The prompts can be information extracted from expert opinions that represents their core semantic content, and are key to guiding the large language model to generate updated questionnaires. Core viewpoints are complete sentences or assertions with clear arguments in the expert opinions; keywords are terms or phrases that highly summarize the theme of the expert opinions.

[0065] The weight value is used to quantify the relative importance or degree of controversy of the prompt information among all expert opinions, and can be derived based on statistical analysis of all expert opinions.

[0066] Logical flaws refer to errors in reasoning, missing evidence, contradictions, or unverified implicit assumptions identified through analysis of preliminary consensus and expert opinions.

[0067] When updating a questionnaire, the prompts and their corresponding weights can be determined. Logical flaws can also be identified based on preliminary consensus and expert opinions. The questionnaire can then be updated based on the prompts, weights, and logical flaws.

[0068] For example, the core argument could be: "Solid-state batteries are the ultimate solution to the range anxiety of electric vehicles." Keywords could include solid-state batteries, smart cockpits, etc. The initial consensus is: "Solid-state batteries are the ultimate solution to the range anxiety of electric vehicles." However, multiple expert opinions include "the R&D cost of solid-state batteries is high" and "the market size is unclear." Therefore, there is a logical flaw between the initial consensus and the expert opinions, which can be interpreted as: "There is a contradiction between the advantages and risks of solid-state batteries."

[0069] Once the aforementioned prompts, corresponding weight values, and logical flaws are identified, a comprehensive analysis of the input information is performed based on a large language model to obtain an updated questionnaire.

[0070] Large language models can be used to uncover the intrinsic connections between core viewpoints, keywords, and logical flaws in expert opinions, thereby accurately generating updated questionnaires.

[0071] Optionally, the prompt information includes core viewpoints; multiple prompt information messages are determined based on the expert opinions, and the weight values ​​corresponding to the prompt information messages include:

[0072] For any expert opinion, the corresponding core viewpoint is determined using the large language model.

[0073] For any core viewpoint, a corresponding first weight value is determined; the first weight value is inversely proportional to the number of core viewpoints.

[0074] Perform a deduplication operation on all core viewpoints to obtain the deduplicated core viewpoints.

[0075] Optionally, for any given expert opinion, a large language model can be used to output the core viewpoint. After determining the core viewpoints of all expert opinions, a first weight value can be determined for each core viewpoint. The first weight value is inversely proportional to the frequency of the core viewpoint's occurrence; that is, the more unique and less frequently mentioned the viewpoint by experts, the higher the first weight value. After determining the core viewpoints of each expert opinion, all core viewpoints can be deduplicated to obtain a list of core viewpoints with their first weight values.

[0076] Optionally, when generating the core viewpoints of expert opinions based on a large language model, prompt words can be used. Specifically, prompt words can be constructed to guide the large language model in completing the summarization task. For example, the prompt word could be: "Please summarize the core viewpoint of the following expert opinion in a very concise and complete declarative sentence."

[0077] When determining the first weight value for a particular core viewpoint, one can determine the number of times that core viewpoint appears among all core viewpoints, and use the reciprocal of that number of occurrences as the corresponding first weight value. For example, if core viewpoint A is extracted by five experts, then the first weight value is 0.2; if core viewpoint 2 is extracted by two experts, then the first weight value is 0.5.

[0078] In addition to the core viewpoints, expert opinions can be analyzed in a stratified manner to identify the main and secondary arguments, as well as their corresponding weight values, thereby obtaining hierarchical information to determine when to update the questionnaire.

[0079] By extracting each core viewpoint and its primary weight, we can focus on unique and potentially forward-looking perspectives to improve the accuracy of updated questionnaires.

[0080] Optionally, the prompt information includes keywords; multiple prompt information messages are determined based on the expert opinions, and the weight values ​​corresponding to the prompt information messages include:

[0081] For any expert opinion, the corresponding keywords are determined using the large language model.

[0082] For any given keyword, a corresponding second weight value is determined; the second weight value is inversely proportional to the number of keywords.

[0083] Perform a deduplication operation on all keywords to obtain the deduplicated keywords.

[0084] Similarly, a large language model can be used to output the results when determining each keyword and its corresponding second weight value. After determining the keywords for all expert opinions, the second weight value for each keyword can also be determined. The second weight value is inversely proportional to the frequency of the keyword's occurrence; that is, the more unique the keyword and the less frequently it is mentioned by experts, the higher its weight. After determining the keywords for each expert opinion, all keywords can be deduplicated to obtain a keyword list with second weight values.

[0085] When determining the second weight value for a specific keyword, you can determine the number of times that keyword appears among all keywords, and use the reciprocal of that number of occurrences as the corresponding second weight value. For example, if keyword A is extracted by five experts, then the second weight value is 0.2.

[0086] By extracting each keyword and its secondary weight value, we can focus on keywords that appear less frequently, thereby improving the accuracy of updating the questionnaire.

[0087] Optionally, identifying logical flaws through the expert opinions includes:

[0088] A preliminary consensus was reached based on the aforementioned expert opinions;

[0089] The initial consensus is then reverse-engineered to identify the logical flaws.

[0090] When identifying logical flaws, we can first determine the preliminary consensus among the experts in this round. For example, we can input the expert opinions and prompts into a large language model, allowing the model to output a preliminary consensus. A preliminary consensus refers to a situation where, after a consensus assessment, no consensus has been reached, but a preliminary consensus exists—that is, a preliminary consensus reached by most experts.

[0091] For example, based on the initial questionnaire survey, the preliminary consensus could be: "Solid-state batteries are the inevitable trend of the next generation of power batteries, which will completely solve the problems of safety and driving range."

[0092] After establishing a preliminary consensus, back-reasoning prompts can be constructed to identify logical flaws. By introducing automatic back-reasoning, the large language model can perform reverse analysis on the preliminary consensus, identify potential logical flaws or inconsistencies, and make targeted corrections in subsequent questionnaires.

[0093] By identifying logical flaws, we can ensure that the final consensus conclusion is logically sound, thus improving the quality of the consensus conclusion.

[0094] Optionally, the updated questionnaire may be sent to each expert, including:

[0095] The updated questionnaire, along with deduplicated keywords, is sent to each expert; the deduplicated keywords are used by the experts to complete the updated questionnaire.

[0096] When sending updated questionnaires to various experts, deduplicated keywords can also be sent simultaneously, allowing experts to complete the updated questionnaires based on these keywords. The deduplicated keywords can be displayed anonymously to all experts to ensure independent judgment and avoid the influence of groupthink.

[0097] Optionally, the deduplicated keywords are extracted from expert opinions and then processed by a large language model to remove duplicate, redundant, or off-topic words. When experts complete the updated questionnaire, they can refer to this list to focus on core issues, avoid feedback bias, and consider keywords that were not previously considered.

[0098] By sending deduplicated keywords to various experts, the accuracy and efficiency of questionnaire completion can be improved, the consensus can be formed more quickly, and the reliability of the consensus results can be enhanced.

[0099] Figure 3 This is a schematic diagram of a complete consensus decision-making method provided by an embodiment of the present invention, as shown below. Figure 3As shown, initially, the system can configure the expert list, email addresses, and questionnaire content to generate an initial questionnaire, which is then sent to each expert via email. It can also receive expert feedback via email, processing the email content into a standardized text format to obtain expert opinions. Furthermore, it can process the expert opinions based on a large language model to obtain an updated questionnaire. In addition, it can perform a consistency assessment of expert opinions to determine if they are in agreement. If not, an updated questionnaire is sent to each expert via email address to collect further expert opinions. This process is repeated multiple times until all expert opinions are consistent, at which point the process ends, and a consensus conclusion is output. Consistency among expert opinions means that the experts have reached a basic agreement.

[0100] By employing automated text analysis and intelligent questionnaire management based on a large language model, manual intervention is significantly reduced, and the speed of expert opinion processing is improved. Furthermore, the objectivity of the research results is enhanced, as the large language model, based on data-driven analysis, avoids subjective biases inherent in human induction. Moreover, this method supports multi-dimensional analysis, including core viewpoints, keywords, and logical flaws, thereby achieving a deeper level of expert consensus. Finally, the adaptive capability of the large language model can be continuously optimized as research progresses, thus improving research efficiency.

[0101] Figure 4 This is a schematic diagram of a consensus decision-making device 40 provided in an embodiment of the present invention. The device includes:

[0102] The sending module 401 is used to send the initial questionnaire to each expert based on the user's input information; the input information includes the email address of each expert and the content of the questionnaire.

[0103] Evaluation module 402 is used to evaluate the consistency of the received expert opinions;

[0104] The processing module 403 is used to determine an updated questionnaire based on a large language model when no consensus is reached, and to send the updated questionnaire to each expert until the opinions of each expert reach a consensus and a consensus conclusion is obtained.

[0105] Optionally, when performing a consistency assessment on the received expert opinions, the evaluation module 402 is specifically used for:

[0106] For any expert opinion, determine the vector corresponding to the expert opinion based on the large language model;

[0107] Calculate the cosine similarity between any two vectors, and calculate the average of all cosine similarities;

[0108] When the average value is greater than the preset value, it is determined that the opinions of the experts have reached a consensus; when the average value is less than the preset value, it is determined that the opinions of the experts have not reached a consensus.

[0109] Optionally, when determining the updated questionnaire based on the large language model, processing module 403 is specifically used for:

[0110] Based on the expert opinions, multiple prompts and their corresponding weight values ​​are determined.

[0111] Logical flaws were identified based on the expert opinions mentioned above;

[0112] The multiple prompts, their corresponding weight values, and the logical flaws are input into the large language model to obtain the updated questionnaire.

[0113] Optionally, the prompt information includes core viewpoints; when the processing module 403 determines multiple prompt information and the weight values ​​corresponding to the prompt information based on the expert opinions, it is specifically used for:

[0114] For any expert opinion, the corresponding core viewpoint is determined using the large language model.

[0115] For any core viewpoint, a corresponding first weight value is determined; the first weight value is inversely proportional to the number of core viewpoints.

[0116] Perform a deduplication operation on all core viewpoints to obtain the deduplicated core viewpoints.

[0117] Optionally, the prompt information includes keywords; when the processing module 403 determines multiple prompt information and the weight values ​​corresponding to the prompt information based on the expert opinions, it is specifically used for:

[0118] For any expert opinion, the corresponding keywords are determined using the large language model.

[0119] For any given keyword, a corresponding second weight value is determined; the second weight value is inversely proportional to the number of keywords.

[0120] Perform a deduplication operation on all keywords to obtain the deduplicated keywords.

[0121] Optionally, when the processing module 403 identifies a logical vulnerability based on the expert opinion, it is specifically used for:

[0122] A preliminary consensus was reached based on the aforementioned expert opinions;

[0123] The initial consensus is then reverse-engineered to identify the logical flaws.

[0124] Optionally, when sending the updated questionnaire to each expert, the processing module 403 is specifically used for:

[0125] The updated questionnaire, along with deduplicated keywords, is sent to each expert; the deduplicated keywords are used by the experts to complete the updated questionnaire.

[0126] The consensus decision-making device 40 provided in this embodiment of the invention can achieve the above-mentioned... Figure 1 The consensus decision-making method shown is similar in principle and technical effect, and will not be described in detail here.

[0127] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Figure 5 As shown, the electronic device provided in this embodiment includes at least one processor 501 and a memory 502. The processor 501 and the memory 502 are connected via a bus 503.

[0128] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to execute the method in the above method embodiment.

[0129] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0130] In the above Figure 5 In the illustrated embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0131] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage.

[0132] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0133] This invention also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method described in the above embodiments.

[0134] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the above method embodiments.

[0135] The aforementioned computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0136] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0137] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0138] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0139] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0140] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0141] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0142] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0143] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A consensus decision-making method, characterized in that, include: Initial questionnaires are sent to various experts based on the user's input information; the input information includes the email addresses of each expert and the content of the questionnaire. A consistency assessment was conducted on the opinions received from all experts. When no consensus is reached, an updated questionnaire is determined based on the large language model and sent to each expert until the opinions of all experts reach a consensus and a consensus conclusion is reached.

2. The method according to claim 1, characterized in that, A consensus assessment was conducted on the received expert opinions, including: For any expert opinion, determine the vector corresponding to the expert opinion based on the large language model; Calculate the cosine similarity between any two vectors, and calculate the average of all cosine similarities; When the average value is greater than the preset value, it is determined that the opinions of the experts have reached a consensus; when the average value is less than the preset value, it is determined that the opinions of the experts have not reached a consensus.

3. The method according to claim 1, characterized in that, The updated questionnaire is determined based on a large language model, including: Based on the expert opinions, multiple prompts and their corresponding weight values ​​are determined. Logical flaws were identified based on the expert opinions mentioned above; The multiple prompts, their corresponding weight values, and the logical flaws are input into the large language model to obtain the updated questionnaire.

4. The method according to claim 3, characterized in that, The prompts include core viewpoints; multiple prompts are determined based on the expert opinions, and the weight values ​​corresponding to the prompts include: For any expert opinion, the corresponding core viewpoint is determined using the large language model. For any core viewpoint, a corresponding first weight value is determined; the first weight value is inversely proportional to the number of core viewpoints. Perform a deduplication operation on all core viewpoints to obtain the deduplicated core viewpoints.

5. The method according to claim 3, characterized in that, The prompt information includes keywords; based on the expert opinions, multiple prompt information messages are determined, and the weight values ​​corresponding to the prompt information messages include: For any expert opinion, the corresponding keywords are determined using the large language model. For any given keyword, a corresponding second weight value is determined; the second weight value is inversely proportional to the number of keywords. Perform a deduplication operation on all keywords to obtain the deduplicated keywords.

6. The method according to claim 3, characterized in that, Logical flaws were identified based on the expert opinions, including: A preliminary consensus was reached based on the aforementioned expert opinions; The initial consensus is then reverse-engineered to identify the logical flaws.

7. The method according to claim 5, characterized in that, The updated questionnaire was sent to the various experts, including: The updated questionnaire, along with deduplicated keywords, is sent to each expert; the deduplicated keywords are used by the experts to complete the updated questionnaire.

8. A consensus decision-making device, characterized in that, include: The sending module is used to send the initial questionnaire to various experts based on the user's input information; The input information includes the email addresses of each expert and the content of the questionnaire; The evaluation module is used to assess the consistency of the received expert opinions. The processing module is used to determine an updated questionnaire based on a large language model when no consensus is reached, and to send the updated questionnaire to each expert until the opinions of all experts reach a consensus and a consensus conclusion is obtained.

9. An electronic device, characterized in that, include: At least one processor and memory; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the method as described in any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images