A Method for Predicting the Evolution of Online Public Opinion with the Involvement of AI Agents and Online Trolls

Abstract

This application relates to a method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls. The method includes: acquiring social media user data; constructing a weighted directed social network containing ordinary netizens, online trolls, and commentary-type AI agents and identifying node types; mapping text content to opinion values, setting the opinion values ​​of online trolls to fixed values, and setting the commentary-type AI agents to generate targeted comment opinion values ​​for different target users; calculating the individual confidence threshold of ordinary netizen nodes, and performing opinion updates based on the extended Deffuant-Weisbuch dynamics equation; iteratively updating until convergence, and outputting the final opinion distribution as the prediction result. This invention, by introducing multi-agent heterogeneous interaction and human intervention, solves the problem of existing models' inadequate characterization of complex public opinion environments, significantly improving the accuracy and modeling completeness of online public opinion evolution prediction.

Description

Technical Field

[0001] This application relates to the field of network information processing technology, and in particular to a method for predicting the evolution of online public opinion under the intervention of AI intelligent agents and online trolls. Background Technology

[0002] With the widespread use of the internet and social media, online public opinion has an increasingly significant impact on social governance, public decision-making, and business activities. Accurate modeling and prediction of the evolution of online public opinion are crucial for understanding public opinion trends and mitigating risks.

[0003] Existing modeling and prediction technologies for the evolution of online public opinion primarily focus on the changing patterns of individual viewpoints during interactions, typically assuming ordinary internet users as the main participants and emphasizing the mechanisms of mutual influence among them. However, these existing solutions neglect the widespread presence of online trolls and artificial intelligence (AI) agents in the real-world online environment.

[0004] In the actual process of public opinion dissemination, online trolls typically influence public opinion through organized and purposeful information intervention; while AI agents, such as intelligent programs that can automatically generate content, can deeply participate in information dissemination by generating comments and summaries. The participation of these entities makes the evolution of public opinion exhibit a more complex multi-agent interactive characteristic, which is fundamentally different from the ideal model of interaction only by ordinary netizens.

[0005] Furthermore, with the development of AI technology, existing AI agents have diversified, mainly including commentary AI and summarizing AI. Different types of AI agents differ significantly in their information generation methods, dissemination mechanisms, and the paths through which they influence user cognition. Ignoring the influence of online trolls and different types of AI agents will lead to a serious deviation between public opinion evolution models and real-world scenarios, thereby reducing the accuracy and reliability of prediction results.

[0006] Therefore, there is an urgent need in related technologies to achieve a more refined characterization of the evolution process of complex online public opinion systems and to improve the accuracy of public opinion evolution prediction and the completeness of modeling. Summary of the Invention

[0007] Therefore, it is necessary to provide a method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls to address the above-mentioned technical problems.

[0008] Firstly, this application provides a method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls. The method includes: Acquire user data from social media platforms, including user registration information, user relationship data, and user historical behavior data; A weighted directed social network is constructed based on the user data and node types are identified. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data. The text content published by ordinary netizens and online trolls is mapped to opinion values ​​within a continuous interval. The opinion values ​​of the online trolls are set to be fixed values ​​that remain unchanged during the evolution of public opinion. The comment-type AI agent is set to generate targeted comment opinion values ​​for different target users it connects to. Calculate the individual confidence threshold of the ordinary internet user node, and perform opinion updates of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term; The opinion update is iteratively executed until the opinion state of the ordinary netizens' nodes reaches the convergence condition, and the final opinion distribution of the ordinary netizens' nodes after convergence is output as the prediction result of public opinion evolution.

[0009] Optionally, in one embodiment of this application, setting the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to includes: Comment opinions are obtained by randomly sampling from a preset interval according to a probability distribution. The comment opinion value is calculated based on the target user's current opinion value.

[0010] Optionally, in one embodiment of this application, calculating the individual confidence threshold of the ordinary internet user node includes: Trust level is calculated based on the number of historical support behaviors of the ordinary internet user node to the neighbor node and the total number of interactions, and the individual confidence threshold is calculated based on the trust level and the edge weight.

[0011] Optionally, in one embodiment of this application, the method further includes: A summarizing AI agent is introduced to generate summary opinions. The summary opinions are either global summary opinions calculated based on the opinion values ​​of all nodes in the weighted directed social network, or highly connected summary opinions calculated based on the opinion values ​​of highly connected nodes in the weighted directed social network.

[0012] Optionally, in one embodiment of this application, the summarizing AI agent's influence on the ordinary internet user node is executed after the viewpoint update based on the extended Deffuant-Weisbuch dynamic equation is completed. As a secondary update to the viewpoint value of the ordinary internet user node, the summarizing AI agent is configured as a module that can be independently turned on or off.

[0013] Optionally, in one embodiment of this application, the public opinion evolution prediction result further includes an opinion polarization index calculated based on the final opinion distribution, the opinion polarization index being used to reflect the degree to which individual opinions deviate from the overall average opinion value.

[0014] Optionally, in one embodiment of this application, the method further includes: Multiple comparative simulation scenarios were constructed, and the final distribution of opinions under the intervention of different subjects was recorded by controlling variables. The intervention effect index was calculated, which was used to quantify the degree of influence of different types of subjects on the evolution of public opinion.

[0015] Secondly, this application also provides a device for predicting the evolution of online public opinion under the intervention of AI agents and online trolls. The device includes: The data acquisition module is used to acquire user data from social media platforms, including user registration information, user relationship data, and user historical behavior data. The network construction and identification module is used to construct a weighted directed social network based on the user data and identify node types. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data. The opinion and comment configuration module is used to map the text content published by ordinary netizens and online trolls to opinion values ​​within a continuous range, set the opinion value of the online trolls to a fixed value that remains unchanged during the evolution of public opinion, and set the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to; The evolutionary computation module is used to calculate the individual confidence threshold of the ordinary internet user node and perform opinion updates of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term; The output module is used to iteratively execute the opinion update until the opinion state of the ordinary netizens' nodes reaches the convergence condition, and output the final opinion distribution of the ordinary netizens' nodes after convergence as the prediction result of public opinion evolution.

[0016] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the steps of the methods described in the various embodiments above.

[0017] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the methods described in the various embodiments above.

[0018] Compared with existing technologies, the aforementioned method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls, by introducing online trolls, commentary-type AI agents, and summarizing AI agents, and combining individual heterogeneity confidence thresholds, extends the modeling of the evolution of public opinion in complex online environments, thus having the following beneficial effects: First, by introducing online trolls as exogenous intervention agents and comment-type AI agents as targeted comment generation agents into the public opinion evolution model, we break through the existing modeling methods that only consider homogeneous interactions between ordinary users. This allows the model to simultaneously represent two types of mechanisms: natural user interaction and external information intervention. As a result, the public opinion evolution equation can simultaneously reflect the process of convergence of opinions among ordinary netizens, the continuous influence of online trolls, and the targeted influence of comment-type AI agents, thereby improving the model's ability to characterize complex public opinion dissemination scenarios.

[0019] Second, a targeted comment generation mechanism is proposed for comment-oriented AI agents. This mechanism does not simplify the comment-oriented AI agent into a single, fixed viewpoint, but instead defines it to generate different comment viewpoint values ​​for different target users. Compared to existing technologies that uniformly treat all propagation agents as node state update objects, this invention extends the AI ​​agent's intervention from the internal state of nodes to a targeted influence on edges. Theoretically, this can more accurately describe the asymmetric propagation phenomena of differentiated push notifications, targeted comments, and pandering generation for different users in real-world networks. Because this targeted comment mechanism directly affects the update equation of the target user, it can improve the model's ability to distinguish the AI ​​agent's intervention path, intervention object, and intervention intensity.

[0020] Third, this invention introduces a confidence threshold calculation mechanism based on historical interaction behavior for ordinary internet users, transforming the confidence threshold from a uniform constant to an individualized parameter. Compared to using a uniform confidence threshold or manually set fixed thresholds, this invention incorporates users' historical support behavior, relationship weights, and interaction differences into the threshold calculation process, enabling a quantitative representation of the differences in opinion tolerance among different ordinary internet users. This directly improves the model's accuracy in characterizing individual behavioral heterogeneity and reduces the bias introduced by the uniform threshold assumption in the evolution results.

[0021] Fourth, this invention further introduces a summarizing AI agent and sets up two modes of operation: a global summarizing strategy and a highly connected summarizing strategy. The newly added macro-information aggregation mechanism enables the model to reflect not only the impact of local neighbor interactions on viewpoints but also the secondary impact process triggered by global summary information or key node summary information. In particular, by adding a summary impact step after the update of ordinary netizens' viewpoints at each time step, this invention establishes a closed-loop structure at the model level from local propagation to global aggregation and then to re-influence, thereby improving the modeling ability for the content propagation effects of platform summaries, hot topic summaries, and information aggregation AI agents.

[0022] Fifth, the summarizing AI agent is set as a configurable module that can be selectively enabled or disabled, and the commenting AI agent and the summarizing AI agent can be intervened independently. This allows the invention to construct multiple comparative simulation scenarios under the same experimental conditions, including scenarios without AI agent intervention, scenarios with only online troll intervention, scenarios with only commenting AI agents, scenarios with only summarizing AI agents, and scenarios with multiple agents involved simultaneously. By recording the final opinion distribution, polarization index, convergence speed, and changes in group structure under different scenarios using the controlled variable method, the differences in the effects of different types of intervention factors on the evolution of public opinion can be directly identified, thereby improving the analyzability of the model. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls in one embodiment. Figure 2 This is a structural block diagram of a network public opinion evolution prediction device with the intervention of AI intelligent agents and online water army in one embodiment; Figure 3 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] In one embodiment, such as Figure 1 As shown, a method for predicting the evolution of online public opinion with the intervention of AI agents and online trolls is provided, including the following steps: S101: Obtain user data from social media platforms, including user registration information, user relationship data, and user historical behavior data.

[0026] In this embodiment, the dataset required for modeling is first obtained from a social media platform, such as Weibo, collected through web crawling. The dataset should include user registration information, user follow relationships, and historical user behavior data, including but not limited to user-posted content, comments, reposts, and likes. The acquired data undergoes preprocessing, including removing outliers, filling in missing values, and structuring the data to form a standardized dataset for subsequent modeling.

[0027] S102: Construct a weighted directed social network based on the user data and identify node types. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data.

[0028] In this embodiment of the application, a weighted directed social network is constructed based on the above data to characterize the relationship structure between users. Let the social network be represented as... , where the set of nodes This refers to individual users participating in the dissemination of public opinion, while also representing a collection of data. Represents the social connections between users, a set of weights. This indicates the strength of the relationship between users, where Indicates user With users The weights of the relationships between them.

[0029] In this social network, the nodes mainly consist of three categories: ordinary netizens, online trolls, and commenting AI agents. The strength of relationships between users varies significantly; therefore, edge weights are characterized by a weighted fusion of multi-dimensional features. Specifically, the relationship weights between users... Defined as:

[0030] in, Metrics representing the attention relationship between users; Indicates the similarity between user registration information; Indicates the intensity of historical interactions between users; Let be the weight coefficient, and satisfy... .

[0031] Furthermore, to characterize the differentiated roles of various entities in the evolution of public opinion, the nodes are categorized and classified. This classification is based on user registration information and behavioral characteristics. In one embodiment, the user registration information includes indicators such as account registration time, real-name authentication status and type, number of followers and following, user nickname and avatar information, geographic information, and account activity level.

[0032] Based on the above information, classification rules or models can be constructed to identify user types. For ordinary netizens, their behavioral characteristics include relatively random posting times, high content diversity, and relatively dispersed interaction relationships; for online trolls, their behavioral characteristics include obvious regularity or concentration in posting times, high content repetition, high frequency of participation in specific topics, and abnormal attention and follower structure; for comment-type AI agents, their behavioral characteristics include rapid information generation, continuous reply behavior, and relatively standardized language expression in text content with certain patterned characteristics.

[0033] In the specific implementation process, a discriminant function can be constructed based on the above registration information and behavioral characteristics, or a supervised learning classification model can be adopted to automatically identify the node type, thereby completing the type labeling of multi-subject nodes and providing a foundation for subsequent public opinion evolution modeling.

[0034] S103: Map the text content published by ordinary netizens and online trolls to opinion values ​​within a continuous interval, set the opinion value of the online trolls to a fixed value that remains unchanged during the evolution of public opinion, and set the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to.

[0035] In this embodiment, after constructing a multi-subject weighted social network, the evolution of opinions in online public opinion is modeled based on the social network. First, the opinions of each subject in the network are numerically represented. For ordinary netizens and online trolls, their opinions are mapped from the text content they publish. Specifically, let users... In time The text set is A pre-trained language model is used to extract semantic features from the text to obtain a semantic representation of the text. Preferably, a sentence vector text representation model is used to encode the text, and its semantic similarity with manually selected extreme viewpoint reference texts is calculated, thereby constructing a viewpoint mapping function. Mapping semantic representations to intervals To obtain user opinion values:

[0036] Wherein, the function The training process employs supervised learning, during which text samples of extremely positive and extremely negative viewpoints on a specific topic are manually selected and assigned values ​​accordingly. and This enables the mapping of the textual semantic space to a continuous viewpoint space.

[0037] For online trolls, their viewpoints are set as exogenous fixed variables in the modeling process, meaning they remain unchanged during the evolution process, in order to characterize their continuous intervention in the spread of public opinion.

[0038] For comment-oriented AI agents, instead of setting a uniform node opinion state, they generate targeted comments based on the different users they are connected to. Specifically, a comment-oriented AI agent is defined. In time For users The generated comment opinion value .

[0039] In one embodiment of this application, setting the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to includes: S201: Randomly sample comments and opinions from a preset interval according to a probability distribution.

[0040] S202: Calculate the comment opinion value based on the target user's current opinion value.

[0041] In one embodiment of this application, the comment generation strategy for generating targeted comment opinion values ​​includes a random comment strategy and a synchronous comment strategy: when using the random comment strategy, for any connected user The generated comment opinion values ​​are determined by the range The probability distribution within is obtained by random sampling, that is:

[0042] When a uniform distribution is used, we have:

[0043] When employing a synchronized comment strategy, comments are generated separately for different connected users, and the comment opinions are consistent with or nearly consistent with the target user's current opinion. Specifically, this includes:

[0044] in, For view synchronization functions, in one embodiment, the following may be adopted:

[0045] Or, to further represent it in a perturbation-inducing form:

[0046] in, To meet The disturbance term.

[0047] Furthermore, connection weights Used to characterize the probability or frequency of triggering commenting behavior. Define the commenting AI agent's interaction with the user. The probability of triggering a comment is:

[0048] This enables the modeling of differentiated commenting behaviors among users with different connections. Based on the stated probability, random sampling is used to determine the time... The collection of AI agents that actually generate comments .

[0049] S104: Calculate the individual confidence threshold of the ordinary internet user node, and perform opinion update of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term.

[0050] In this embodiment, an individual confidence threshold is introduced during the opinion evolution process to characterize the differences in an individual's acceptance of different viewpoints. In this invention, the confidence threshold is modeled only for ordinary internet user nodes. After determining the initial states of ordinary internet users, online trolls, and comment-based AI agents, the process is first based on the stated time... The opinion state determines the current opinion value of ordinary netizens in the network. Let ordinary netizens be... In time opinion value Its set of neighboring nodes is Based on the weighted social network and individual confidence thresholds, a model is constructed to depict the opinion interaction process among ordinary internet users.

[0051] In one embodiment, the Deffuant-Weisbuch dynamic equation is used to describe the view update process of ordinary internet user nodes. When an ordinary internet user node... Its neighboring nodes When an exchange of opinions occurs, the first step is to determine whether the difference between the two opinions is within a confidence threshold, i.e.:

[0052] in, Represents ordinary internet user nodes The individual confidence threshold. When the above conditions are met, it indicates that the node... The viewpoint is at the node Within the acceptable range, then the node The next time step will be updated as follows:

[0053] in, Denotes the convergence coefficient of the viewpoints, and ; Represents a node For nodes Relationship weights are used to characterize the strength of interactive influence.

[0054] When ordinary netizens node When multiple neighboring nodes interact effectively within the same time step, it can be further represented as a weighted aggregation update:

[0055] in, Indicates time With nodes The set of neighboring nodes that satisfy the confidence constraint, i.e.:

[0056] This enables dynamic updates of ordinary netizens' opinions based on social network structure, node relationship strength, and individual heterogeneity confidence threshold.

[0057] Furthermore, to characterize the impact of comment-based AI agents on the evolution of ordinary netizens' opinions, a comment intervention item is introduced during the process of updating ordinary netizens' opinions. Let's assume a time frame... Targeting ordinary netizens The set of AI agents that generate comments is The corresponding targeted comment opinion value Then, considering comment intervention, the formula for updating the opinions of ordinary netizens can be expressed as:

[0058] in, This represents the influence coefficient of the commenting AI agent. Commentary AI agent For ordinary netizens Comments affect weight.

[0059] During the aforementioned update process, the opinion value of online troll nodes remains unchanged, and they do not participate in updating their own opinions. However, as neighboring nodes, they can influence ordinary netizens through the aforementioned interaction formula. Therefore, the Deffuant-Weisbuch dynamic equation is extended in this invention into a multi-agent opinion evolution equation that simultaneously considers the combined effects of ordinary netizens, online trolls, and commentary-based AI agents.

[0060] In one embodiment of this application, calculating the individual confidence threshold of the ordinary internet user node includes: Trust level is calculated based on the number of historical support behaviors of the ordinary internet user node to the neighbor node and the total number of interactions, and the individual confidence threshold is calculated based on the trust level and the edge weight.

[0061] In one embodiment of this application, a regular internet user node is set. The confidence threshold is It estimates based on historical interaction behavior. Specifically, let the user... For users The number of historical support behaviors is The total number of interactions is Trust level is then defined as:

[0062] in, For smoothing parameters.

[0063] Furthermore, define the user The confidence threshold is:

[0064] in, Based on the confidence level, To adjust the parameters.

[0065] The above methods enable the modeling of opinion representation and individual heterogeneity confidence thresholds under multi-agent conditions, providing a foundation for subsequent opinion interaction and evolution processes.

[0066] In one embodiment of this application, the method further includes: A summarizing AI agent is introduced to generate summary opinions. The summary opinions are either global summary opinions calculated based on the opinion values ​​of all nodes in the weighted directed social network, or highly connected summary opinions calculated based on the opinion values ​​of highly connected nodes in the weighted directed social network.

[0067] In one embodiment of this application, the summarizing AI agent's influence on the ordinary internet user node is executed after the viewpoint update based on the extended Deffuant-Weisbuch dynamic equation is completed. As a secondary update to the viewpoint value of the ordinary internet user node, the summarizing AI agent is configured as a module that can be independently turned on or off.

[0068] In one embodiment of this application, after the general public opinion is updated at each time step, a summarizing AI agent can be selectively introduced to further analyze the general public opinion to assess the impact of the summary information on the evolution of public opinion. The summarizing AI agent is a configurable module that can be turned on or off as needed during system operation.

[0069] When a summary-type AI agent is enabled, in one embodiment, the summary generation strategy includes two types: a global summary strategy and a highly interconnected summary strategy.

[0070] When a global summarization strategy is adopted, an overall summary viewpoint is generated based on the current viewpoint state of each node in the entire social network. Let the set of all nodes in the network be denoted as . Then the global summary viewpoint value It can be represented as:

[0071] in, Represents a node The contribution weight when generating the global summary. In one embodiment, It can be taken as node activity, node influence, or a preset combination of weights.

[0072] When employing a highly connected summarizing strategy, the first step is to identify nodes in the social network with a large number of connected users, forming a highly connected node set. Then, a summary viewpoint is generated based on the viewpoint status in the set of highly connected nodes. Correspondingly, the value of the highly connected summary viewpoint... It can be represented as:

[0073] in, Indicates a high-connectivity node Contribution weights in summary generation. In one embodiment, highly connected nodes are selected based on their node degree, weighted degree, or centrality index exceeding a preset threshold.

[0074] After the summarizing AI agent generates a summary viewpoint, it further influences ordinary internet user nodes. Let the value of the summary viewpoint be denoted as... It can be selected based on the global summary strategy or the high-connectivity summary strategy, then the ordinary user node The update form after incorporating the effect of the summary can be expressed as:

[0075] in, This indicates that the summary information is relevant to ordinary internet users. The influence intensity coefficient. In one embodiment, It can be set to a constant, or it can be based on ordinary internet user nodes. Individualize information preferences, activity levels, or historical acceptance of summary information.

[0076] To ensure the flexibility of system operation, when the summary-type AI agent module is turned off, the process of generating summary viewpoints and updating impact is not executed; when the module is turned on, the summary impact calculation is performed after the update of ordinary netizens' viewpoints at each time step, thereby enabling a comparative analysis of the public opinion evolution results under the two scenarios of whether or not the summary-type AI agent intervenes.

[0077] After updating the viewpoint at each time step, the iteration continues to the next time step until the viewpoint state of ordinary netizens' nodes in the network reaches the convergence condition. In one embodiment, the convergence condition can be expressed as:

[0078] in, Represents a set of ordinary internet user nodes. This indicates a preset convergence threshold. When the above conditions are met, the evolution of online public opinion is considered to have reached a stable state, and the final opinion distribution result is output.

[0079] Through the above methods, we have achieved dynamic evolution modeling and iterative prediction of public opinion in multi-subject weighted social networks, combining individual heterogeneity confidence thresholds, commentary AI agent intervention, and optional intervention mechanisms of summarizing AI agents.

[0080] S105: Iteratively execute the opinion update until the opinion state of the ordinary netizens' nodes reaches the convergence condition, and output the final opinion distribution of the ordinary netizens' nodes after convergence as the prediction result of public opinion evolution.

[0081] In one embodiment of this application, the public opinion evolution prediction result further includes an opinion polarization index calculated based on the final opinion distribution, the opinion polarization index being used to reflect the degree to which individual opinions deviate from the overall average opinion value.

[0082] In one embodiment of this application, after completing the iterative calculation of the evolution of online public opinion and reaching a convergence state, the final opinion distribution results of the ordinary netizens' nodes are extracted and analyzed to form a public opinion evolution prediction result. In one embodiment, the prediction result includes, but is not limited to, the following indicators: the opinion distribution structure in the network, the degree of opinion polarization, the group division, and the opinion status of key nodes.

[0083] Specifically, let the converged set of ordinary internet user nodes be . Their viewpoint is Then, the overall opinion distribution function can be obtained through statistical analysis. This is used to describe the distribution characteristics of viewpoints in a network. Furthermore, to characterize the degree of public opinion polarization, a polarization index can be defined as:

[0084] in, This represents the overall average opinion value. The indicator is used to reflect the degree to which individual opinions deviate from the overall mean, thus characterizing the dispersion or polarization of public opinion.

[0085] Furthermore, ordinary internet user nodes can be grouped based on their opinion value. In one embodiment, by setting an opinion range threshold, nodes are divided into supporting, neutral, and opposing groups, for example:

[0086] in, This is a preset threshold parameter. Based on the above division, the size and structural characteristics of different groups can be further analyzed.

[0087] In one embodiment of this application, the method further includes: Multiple comparative simulation scenarios were constructed, and the final distribution of opinions under the intervention of different subjects was recorded by controlling variables. The intervention effect index was calculated, which was used to quantify the degree of influence of different types of subjects on the evolution of public opinion.

[0088] In one embodiment of this application, to further evaluate the impact of different subjects (online trolls, commentary AI agents, and summarizing AI agents) on the evolution of public opinion, a comparative simulation mechanism is constructed in one embodiment. Specifically, by controlling variables, multiple sets of simulations are conducted under the same initial conditions, including: a baseline scenario without AI agents and online trolls, a scenario with only online trolls, a scenario with commentary AI agents, and a scenario with summarizing AI agents. Let the final distribution of opinions under different scenarios be as follows: Then the intervention effect index can be defined as:

[0089] in, These represent different intervention scenarios. The above indicators can quantify the degree of impact of different types of subject intervention on the evolution of public opinion.

[0090] Furthermore, in one embodiment, the time series during the evolution process can be recorded to obtain the viewpoint evolution trajectory at each time step. This enables dynamic analysis of the evolution path of public opinion, including convergence speed, stage fluctuation characteristics, and identification of key time nodes.

[0091] It should be understood that although the steps in the flowcharts of the embodiments described above 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 flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0092] Based on the same inventive concept, this application also provides an AI agent and internet troll intervention-based network public opinion evolution prediction device for implementing the above-mentioned method for predicting the evolution of online public opinion with the intervention of AI agents and internet trolls. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the AI ​​agent and internet troll intervention-based network public opinion evolution prediction device provided below can be found in the limitations of the AI ​​agent and internet troll intervention-based network public opinion evolution prediction method described above, and will not be repeated here.

[0093] In one embodiment, such as Figure 2 As shown, a network public opinion evolution prediction device 200 with the intervention of AI intelligent agents and online trolls is provided, including: a data acquisition module 201, a network construction and identification module 202, an opinion and comment configuration module 203, an evolution calculation module 204, and an output module 205, wherein: The data acquisition module 201 is used to acquire user data from social media platforms, including user registration information, user relationship data, and user historical behavior data.

[0094] The network construction and identification module 202 is used to construct a weighted directed social network based on the user data and identify the node types. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data.

[0095] The opinion and comment configuration module 203 is used to map the text content published by ordinary netizens and online trolls to opinion values ​​within a continuous range, set the opinion value of the online trolls to a fixed value that remains unchanged during the evolution of public opinion, and set the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to.

[0096] Evolutionary computation module 204 is used to calculate the individual confidence threshold of the ordinary internet user node and perform opinion updates of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term.

[0097] Output module 205 is used to iteratively execute the opinion update until the opinion state of the ordinary netizens node reaches the convergence condition, and output the final opinion distribution of the ordinary netizens node after convergence as the public opinion evolution prediction result.

[0098] In one embodiment of this application, setting the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to includes: Comment opinions are obtained by randomly sampling from a preset interval according to a probability distribution. The comment opinion value is calculated based on the target user's current opinion value.

[0099] In one embodiment of this application, calculating the individual confidence threshold of the ordinary internet user node includes: Trust level is calculated based on the number of historical support behaviors of the ordinary internet user node to the neighbor node and the total number of interactions, and the individual confidence threshold is calculated based on the trust level and the edge weight.

[0100] In one embodiment of this application, the method further includes: A summarizing AI agent is introduced to generate summary opinions. The summary opinions are either global summary opinions calculated based on the opinion values ​​of all nodes in the weighted directed social network, or highly connected summary opinions calculated based on the opinion values ​​of highly connected nodes in the weighted directed social network.

[0101] In one embodiment of this application, the summarizing AI agent's influence on the ordinary internet user node is executed after the viewpoint update based on the extended Deffuant-Weisbuch dynamic equation is completed. As a secondary update to the viewpoint value of the ordinary internet user node, the summarizing AI agent is configured as a module that can be independently turned on or off.

[0102] In one embodiment of this application, the public opinion evolution prediction result further includes an opinion polarization index calculated based on the final opinion distribution, the opinion polarization index being used to reflect the degree to which individual opinions deviate from the overall average opinion value.

[0103] In one embodiment of this application, the method further includes: Multiple comparative simulation scenarios were constructed, and the final distribution of opinions under the intervention of different subjects was recorded by controlling variables. The intervention effect index was calculated, which was used to quantify the degree of influence of different types of subjects on the evolution of public opinion.

[0104] The modules in the aforementioned AI-powered intelligent agent and online troll intervention-based online public opinion evolution prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0105] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for predicting the evolution of online public opinion under the intervention of AI agents and online trolls. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0106] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0107] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0108] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0109] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0110] 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.

[0111] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0112] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0113] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A network public opinion evolution prediction method under the intervention of an AI agent and network water army, characterized in that, The method includes: Acquire user data from social media platforms, including user registration information, user relationship data, and user historical behavior data; A weighted directed social network is constructed based on the user data and node types are identified. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data. The text content published by ordinary netizens and online trolls is mapped to opinion values ​​within a continuous interval. The opinion values ​​of the online trolls are set to be fixed values ​​that remain unchanged during the evolution of public opinion. The comment-type AI agent is set to generate targeted comment opinion values ​​for different target users it connects to. Calculate the individual confidence threshold of the ordinary internet user node, and perform opinion updates of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term; The opinion update is iteratively executed until the opinion state of the ordinary netizens' nodes reaches the convergence condition, and the final opinion distribution of the ordinary netizens' nodes after convergence is output as the prediction result of public opinion evolution.

2. The method according to claim 1, wherein, The configuration of the comment-based AI agent to generate targeted comment opinion values ​​for different target users it connects to includes: Comment opinions are obtained by randomly sampling from a preset interval according to a probability distribution. The comment opinion value is calculated based on the target user's current opinion value.

3. The method according to claim 1, wherein, The calculation of the individual confidence threshold for the ordinary internet user node includes: Trust level is calculated based on the number of historical support behaviors of the ordinary internet user node to the neighbor node and the total number of interactions, and the individual confidence threshold is calculated based on the trust level and the edge weight.

4. The method according to claim 1, wherein, The method further includes: A summarizing AI agent is introduced to generate summary opinions. The summary opinions are either global summary opinions calculated based on the opinion values ​​of all nodes in the weighted directed social network, or highly connected summary opinions calculated based on the opinion values ​​of highly connected nodes in the weighted directed social network.

5. The method according to claim 4, wherein, The summarizing AI agent's influence on the ordinary internet user node is executed after the viewpoint update based on the extended Deffuant-Weisbuch dynamic equation is completed. As a secondary update of the viewpoint value of the ordinary internet user node, the summarizing AI agent is configured as a module that can be independently turned on or off.

6. The method according to claim 1, wherein, The public opinion evolution prediction results also include an opinion polarization index calculated based on the final opinion distribution. The opinion polarization index is used to reflect the degree to which individual opinions deviate from the overall average opinion value.

7. The method according to claim 1, wherein, The method further includes: Multiple comparative simulation scenarios were constructed, and the final distribution of opinions under the intervention of different subjects was recorded by controlling variables. The intervention effect index was calculated, which was used to quantify the degree of influence of different types of subjects on the evolution of public opinion.

8. An AI agent and network water army intervention under the network public opinion evolution prediction device, characterized in that, The device includes: The data acquisition module is used to acquire user data from social media platforms, including user registration information, user relationship data, and user historical behavior data. The network construction and identification module is used to construct a weighted directed social network based on the user data and identify node types. The nodes of the weighted directed social network include ordinary netizens, online trolls, and comment-type AI agents. The edge weights are used to characterize the strength of the relationship between users. The node types are identified based on the user registration information and user historical behavior data. The opinion and comment configuration module is used to map the text content published by ordinary netizens and online trolls to opinion values ​​within a continuous range, set the opinion value of the online trolls to a fixed value that remains unchanged during the evolution of public opinion, and set the comment-type AI agent to generate targeted comment opinion values ​​for different target users it connects to; The evolutionary computation module is used to calculate the individual confidence threshold of the ordinary internet user node and perform opinion updates of the ordinary internet user node based on the extended Deffuant-Weisbuch dynamic equation, wherein the opinion update term includes ordinary internet user neighbor interaction term and commentary AI agent comment intervention term; The output module is used to iteratively execute the opinion update until the opinion state of the ordinary netizens' nodes reaches the convergence condition, and output the final opinion distribution of the ordinary netizens' nodes after convergence as the prediction result of public opinion evolution. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images