Patrick Gerard, Svitlana Volkova, Louis Penafiel, Kristina Lerman, Tim Weninger
Following the Russian Federation's full-scale invasion of Ukraine in February 2022, a multitude of information narratives emerged within both pro-Russian and pro-Ukrainian communities online. As the conflict progresses, so too do the information narratives, constantly adapting and influencing local and global community perceptions and attitudes. This dynamic nature of the evolving information environment (IE) underscores a critical need to fully discern how narratives evolve and affect online communities. Existing research, however, often fails to capture information narrative evolution, overlooking both the fluid nature of narratives and the internal mechanisms that drive their evolution. Recognizing this, we introduce a novel approach designed to both model narrative evolution and uncover the underlying mechanisms driving them. In this work we perform a comparative discourse analysis across communities on Telegram covering the initial three months following the invasion. First, we uncover substantial disparities in narratives and perceptions between pro-Russian and pro-Ukrainian communities. Then, we probe deeper into prevalent narratives of each group, identifying key themes and examining the underlying mechanisms fueling their evolution. Finally, we explore influences and factors that may shape the development and spread of narratives.
{"title":"Modeling Information Narrative Detection and Evolution on Telegram during the Russia-Ukraine War","authors":"Patrick Gerard, Svitlana Volkova, Louis Penafiel, Kristina Lerman, Tim Weninger","doi":"arxiv-2409.07684","DOIUrl":"https://doi.org/arxiv-2409.07684","url":null,"abstract":"Following the Russian Federation's full-scale invasion of Ukraine in February\u00002022, a multitude of information narratives emerged within both pro-Russian and\u0000pro-Ukrainian communities online. As the conflict progresses, so too do the\u0000information narratives, constantly adapting and influencing local and global\u0000community perceptions and attitudes. This dynamic nature of the evolving\u0000information environment (IE) underscores a critical need to fully discern how\u0000narratives evolve and affect online communities. Existing research, however,\u0000often fails to capture information narrative evolution, overlooking both the\u0000fluid nature of narratives and the internal mechanisms that drive their\u0000evolution. Recognizing this, we introduce a novel approach designed to both\u0000model narrative evolution and uncover the underlying mechanisms driving them.\u0000In this work we perform a comparative discourse analysis across communities on\u0000Telegram covering the initial three months following the invasion. First, we\u0000uncover substantial disparities in narratives and perceptions between\u0000pro-Russian and pro-Ukrainian communities. Then, we probe deeper into prevalent\u0000narratives of each group, identifying key themes and examining the underlying\u0000mechanisms fueling their evolution. Finally, we explore influences and factors\u0000that may shape the development and spread of narratives.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"8 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226899","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Avijit Ghosh, Tomo Lazovich, Kristian Lum, Christo Wilson
Many existing fairness metrics measure group-wise demographic disparities in system behavior or model performance. Calculating these metrics requires access to demographic information, which, in industrial settings, is often unavailable. By contrast, economic inequality metrics, such as the Gini coefficient, require no demographic data to measure. However, reductions in economic inequality do not necessarily correspond to reductions in demographic disparities. In this paper, we empirically explore the relationship between demographic-free inequality metrics -- such as the Gini coefficient -- and standard demographic bias metrics that measure group-wise model performance disparities specifically in the case of engagement inequality on Twitter. We analyze tweets from 174K users over the duration of 2021 and find that demographic-free impression inequality metrics are positively correlated with gender, race, and age disparities in the average case, and weakly (but still positively) correlated with demographic bias in the worst case. We therefore recommend inequality metrics as a potentially useful proxy measure of average group-wise disparities, especially in cases where such disparities cannot be measured directly. Based on these results, we believe they can be used as part of broader efforts to improve fairness between demographic groups in scenarios like content recommendation on social media.
{"title":"Reducing Population-level Inequality Can Improve Demographic Group Fairness: a Twitter Case Study","authors":"Avijit Ghosh, Tomo Lazovich, Kristian Lum, Christo Wilson","doi":"arxiv-2409.08135","DOIUrl":"https://doi.org/arxiv-2409.08135","url":null,"abstract":"Many existing fairness metrics measure group-wise demographic disparities in\u0000system behavior or model performance. Calculating these metrics requires access\u0000to demographic information, which, in industrial settings, is often\u0000unavailable. By contrast, economic inequality metrics, such as the Gini\u0000coefficient, require no demographic data to measure. However, reductions in\u0000economic inequality do not necessarily correspond to reductions in demographic\u0000disparities. In this paper, we empirically explore the relationship between\u0000demographic-free inequality metrics -- such as the Gini coefficient -- and\u0000standard demographic bias metrics that measure group-wise model performance\u0000disparities specifically in the case of engagement inequality on Twitter. We\u0000analyze tweets from 174K users over the duration of 2021 and find that\u0000demographic-free impression inequality metrics are positively correlated with\u0000gender, race, and age disparities in the average case, and weakly (but still\u0000positively) correlated with demographic bias in the worst case. We therefore\u0000recommend inequality metrics as a potentially useful proxy measure of average\u0000group-wise disparities, especially in cases where such disparities cannot be\u0000measured directly. Based on these results, we believe they can be used as part\u0000of broader efforts to improve fairness between demographic groups in scenarios\u0000like content recommendation on social media.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"97 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142214675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper introduces a fine-grained contrastive learning scheme for unsupervised node clustering. Previous clustering methods only focus on a small feature set (class-dependent features), which demonstrates explicit clustering characteristics, ignoring the rest of the feature spaces (class-invariant features). This paper exploits class-invariant features via graph contrastive learning to discover additional high-quality features for unsupervised clustering. We formulate a novel node-level fine-grained augmentation framework for self-supervised learning, which iteratively identifies competitive contrastive samples from the whole feature spaces, in the form of positive and negative examples of node relations. While positive examples of node relations are usually expressed as edges in graph homophily, negative examples are implicit without a direct edge. We show, however, that simply sampling nodes beyond the local neighborhood results in less competitive negative pairs, that are less effective for contrastive learning. Inspired by counterfactual augmentation, we instead sample competitive negative node relations by creating virtual nodes that inherit (in a self-supervised fashion) class-invariant features, while altering class-dependent features, creating contrasting pairs that lie closer to the boundary and offering better contrast. Consequently, our experiments demonstrate significant improvements in supervised node clustering tasks on six baselines and six real-world social network datasets.
{"title":"Unsupervised node clustering via contrastive hard sampling","authors":"Hang Cui, Tarek Abdelzaher","doi":"arxiv-2409.07718","DOIUrl":"https://doi.org/arxiv-2409.07718","url":null,"abstract":"This paper introduces a fine-grained contrastive learning scheme for\u0000unsupervised node clustering. Previous clustering methods only focus on a small\u0000feature set (class-dependent features), which demonstrates explicit clustering\u0000characteristics, ignoring the rest of the feature spaces (class-invariant\u0000features). This paper exploits class-invariant features via graph contrastive\u0000learning to discover additional high-quality features for unsupervised\u0000clustering. We formulate a novel node-level fine-grained augmentation framework\u0000for self-supervised learning, which iteratively identifies competitive\u0000contrastive samples from the whole feature spaces, in the form of positive and\u0000negative examples of node relations. While positive examples of node relations\u0000are usually expressed as edges in graph homophily, negative examples are\u0000implicit without a direct edge. We show, however, that simply sampling nodes\u0000beyond the local neighborhood results in less competitive negative pairs, that\u0000are less effective for contrastive learning. Inspired by counterfactual\u0000augmentation, we instead sample competitive negative node relations by creating\u0000virtual nodes that inherit (in a self-supervised fashion) class-invariant\u0000features, while altering class-dependent features, creating contrasting pairs\u0000that lie closer to the boundary and offering better contrast. Consequently, our\u0000experiments demonstrate significant improvements in supervised node clustering\u0000tasks on six baselines and six real-world social network datasets.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226896","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Echo chambers and online discourses have become prevalent social phenomena where communities engage in dramatic intra-group confirmations and inter-group hostility. Polarization detection is a rising research topic for detecting and identifying such polarized groups. Previous works on polarization detection primarily focus on hand-crafted features derived from dataset-specific characteristics and prior knowledge, which fail to generalize to other datasets. This paper proposes a unified self-supervised polarization detection framework, outperforming previous methods in unsupervised and semi-supervised polarization detection tasks on various publicly available datasets. Our framework utilizes a dual contrastive objective (DocTra): (1) interaction-level: to contrast between node interactions to extract critical features on interaction patterns, and (2) feature-level: to contrast extracted polarized and invariant features to encourage feature decoupling. Our experiments extensively evaluate our methods again 7 baselines on 7 public datasets, demonstrating significant performance improvements.
{"title":"Polarization Detection on Social Networks: dual contrastive objectives for Self-supervision","authors":"Hang Cui, Tarek Abdelzaher","doi":"arxiv-2409.07716","DOIUrl":"https://doi.org/arxiv-2409.07716","url":null,"abstract":"Echo chambers and online discourses have become prevalent social phenomena\u0000where communities engage in dramatic intra-group confirmations and inter-group\u0000hostility. Polarization detection is a rising research topic for detecting and\u0000identifying such polarized groups. Previous works on polarization detection\u0000primarily focus on hand-crafted features derived from dataset-specific\u0000characteristics and prior knowledge, which fail to generalize to other\u0000datasets. This paper proposes a unified self-supervised polarization detection\u0000framework, outperforming previous methods in unsupervised and semi-supervised\u0000polarization detection tasks on various publicly available datasets. Our\u0000framework utilizes a dual contrastive objective (DocTra): (1)\u0000interaction-level: to contrast between node interactions to extract critical\u0000features on interaction patterns, and (2) feature-level: to contrast extracted\u0000polarized and invariant features to encourage feature decoupling. Our\u0000experiments extensively evaluate our methods again 7 baselines on 7 public\u0000datasets, demonstrating significant performance improvements.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226903","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Kian Ahrabian, Casandra Rusti, Ziao Wang, Jay Pujara, Kristina Lerman
The COVID-19 pandemic profoundly impacted people globally, yet its effect on scientists and research institutions has yet to be fully examined. To address this knowledge gap, we use a newly available bibliographic dataset covering tens of millions of papers and authors to investigate changes in research activity and collaboration during this period. Employing statistical methods, we analyze the pandemic's disruptions on the participation, productivity, and collaborations of researchers at the top 1,000 institutions worldwide based on historical productivity, taking into account variables such as geography, researcher seniority and gender, and field of study. Our findings reveal an unexpected trend: research activity and output significantly increased in the early stages of the pandemic, indicating a surprising resilience in the scientific community. However, by the end of 2022, there was a notable reversion to historical trends in research participation and productivity. This reversion suggests that the initial spike in research activity was a short-lived disruption rather than a permanent shift. As such, monitoring scientific outputs in 2023 and beyond becomes crucial. There may be a delayed negative effect of the pandemic on research, given the long time horizon for many research fields and the temporary closure of wet labs. Further analysis is needed to fully comprehend the factors that underpin the resilience of scientific innovation in the face of global crises. Our study provides an initial comprehensive exploration up to the end of 2022, offering valuable insights into how the scientific community has adapted and responded over the course of the pandemic.
{"title":"Surprising Resilience of Science During a Global Pandemic: A Large-Scale Descriptive Analysis","authors":"Kian Ahrabian, Casandra Rusti, Ziao Wang, Jay Pujara, Kristina Lerman","doi":"arxiv-2409.07710","DOIUrl":"https://doi.org/arxiv-2409.07710","url":null,"abstract":"The COVID-19 pandemic profoundly impacted people globally, yet its effect on\u0000scientists and research institutions has yet to be fully examined. To address\u0000this knowledge gap, we use a newly available bibliographic dataset covering\u0000tens of millions of papers and authors to investigate changes in research\u0000activity and collaboration during this period. Employing statistical methods,\u0000we analyze the pandemic's disruptions on the participation, productivity, and\u0000collaborations of researchers at the top 1,000 institutions worldwide based on\u0000historical productivity, taking into account variables such as geography,\u0000researcher seniority and gender, and field of study. Our findings reveal an\u0000unexpected trend: research activity and output significantly increased in the\u0000early stages of the pandemic, indicating a surprising resilience in the\u0000scientific community. However, by the end of 2022, there was a notable\u0000reversion to historical trends in research participation and productivity. This\u0000reversion suggests that the initial spike in research activity was a\u0000short-lived disruption rather than a permanent shift. As such, monitoring\u0000scientific outputs in 2023 and beyond becomes crucial. There may be a delayed\u0000negative effect of the pandemic on research, given the long time horizon for\u0000many research fields and the temporary closure of wet labs. Further analysis is\u0000needed to fully comprehend the factors that underpin the resilience of\u0000scientific innovation in the face of global crises. Our study provides an\u0000initial comprehensive exploration up to the end of 2022, offering valuable\u0000insights into how the scientific community has adapted and responded over the\u0000course of the pandemic.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142214676","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lukas Hueller, Tim Kuffner, Matthias Schneider, Leo Schuhmann, Virginie Cauderay, Tolga Buz, Vincent Beermann, Falk Uebernickel
Enabling supply chain transparency (SCT) is essential for regulatory compliance and meeting sustainability standards. Multi-tier SCT plays a pivotal role in identifying and mitigating an organization's operational, environmental, and social (ESG) risks. While research observes increasing efforts towards SCT, a minority of companies are currently publishing supply chain information. Using the Design Science Research approach, we develop a collaborative platform for supply chain transparency. We derive design requirements, formulate design principles, and evaluate the artefact with industry experts. Our artefact is initialized with publicly available supply chain data through an automated pipeline designed to onboard future participants to our platform. This work contributes to SCT research by providing insights into the challenges and opportunities of implementing multi-tier SCT and offers a practical solution that encourages organizations to participate in a transparent ecosystem.
{"title":"Designing a Collaborative Platform for Advancing Supply Chain Transparency","authors":"Lukas Hueller, Tim Kuffner, Matthias Schneider, Leo Schuhmann, Virginie Cauderay, Tolga Buz, Vincent Beermann, Falk Uebernickel","doi":"arxiv-2409.08104","DOIUrl":"https://doi.org/arxiv-2409.08104","url":null,"abstract":"Enabling supply chain transparency (SCT) is essential for regulatory\u0000compliance and meeting sustainability standards. Multi-tier SCT plays a pivotal\u0000role in identifying and mitigating an organization's operational,\u0000environmental, and social (ESG) risks. While research observes increasing\u0000efforts towards SCT, a minority of companies are currently publishing supply\u0000chain information. Using the Design Science Research approach, we develop a\u0000collaborative platform for supply chain transparency. We derive design\u0000requirements, formulate design principles, and evaluate the artefact with\u0000industry experts. Our artefact is initialized with publicly available supply\u0000chain data through an automated pipeline designed to onboard future\u0000participants to our platform. This work contributes to SCT research by\u0000providing insights into the challenges and opportunities of implementing\u0000multi-tier SCT and offers a practical solution that encourages organizations to\u0000participate in a transparent ecosystem.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"10 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226898","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lutz Oettershagen, Athanasios L. Konstantinidis, Fariba Ranjbar, Giuseppe F. Italiano
Social network users are commonly connected to hundreds or even thousands of other users. However, these ties are not all of equal strength; for example, we often are connected to good friends or family members as well as acquaintances. Inferring the tie strengths is an essential task in social network analysis. Common approaches classify the ties into strong and weak edges based on the network topology using the strong triadic closure (STC). The STC states that if for three nodes, $textit{A}$, $textit{B}$, and $textit{C}$, there are strong ties between $textit{A}$ and $textit{B}$, as well as $textit{A}$ and $textit{C}$, there has to be a (weak or strong) tie between $textit{B}$ and $textit{C}$. Moreover, a variant of the STC called STC+ allows adding new weak edges to obtain improved solutions. Recently, the focus of social network analysis has been shifting from single-layer to multilayer networks due to their ability to represent complex systems with multiple types of interactions or relationships in multiple social network platforms like Facebook, LinkedIn, or X (formerly Twitter). However, straightforwardly applying the STC separately to each layer of multilayer networks usually leads to inconsistent labelings between layers. Avoiding such inconsistencies is essential as they contradict the idea that tie strengths represent underlying, consistent truths about the relationships between users. Therefore, we adapt the definitions of the STC and STC+ for multilayer networks and provide ILP formulations to solve the problems exactly. Solving the ILPs is computationally costly; hence, we additionally provide an efficient 2-approximation for the STC and a 6-approximation for the STC+ minimization variants. The experiments show that, unlike standard approaches, our new highly efficient algorithms lead to consistent strong/weak labelings of the multilayer network edges.
{"title":"Consistent Strong Triadic Closure in Multilayer Networks","authors":"Lutz Oettershagen, Athanasios L. Konstantinidis, Fariba Ranjbar, Giuseppe F. Italiano","doi":"arxiv-2409.08405","DOIUrl":"https://doi.org/arxiv-2409.08405","url":null,"abstract":"Social network users are commonly connected to hundreds or even thousands of\u0000other users. However, these ties are not all of equal strength; for example, we\u0000often are connected to good friends or family members as well as acquaintances.\u0000Inferring the tie strengths is an essential task in social network analysis.\u0000Common approaches classify the ties into strong and weak edges based on the\u0000network topology using the strong triadic closure (STC). The STC states that if\u0000for three nodes, $textit{A}$, $textit{B}$, and $textit{C}$, there are strong\u0000ties between $textit{A}$ and $textit{B}$, as well as $textit{A}$ and\u0000$textit{C}$, there has to be a (weak or strong) tie between $textit{B}$ and\u0000$textit{C}$. Moreover, a variant of the STC called STC+ allows adding new weak\u0000edges to obtain improved solutions. Recently, the focus of social network\u0000analysis has been shifting from single-layer to multilayer networks due to\u0000their ability to represent complex systems with multiple types of interactions\u0000or relationships in multiple social network platforms like Facebook, LinkedIn,\u0000or X (formerly Twitter). However, straightforwardly applying the STC separately\u0000to each layer of multilayer networks usually leads to inconsistent labelings\u0000between layers. Avoiding such inconsistencies is essential as they contradict\u0000the idea that tie strengths represent underlying, consistent truths about the\u0000relationships between users. Therefore, we adapt the definitions of the STC and\u0000STC+ for multilayer networks and provide ILP formulations to solve the problems\u0000exactly. Solving the ILPs is computationally costly; hence, we additionally\u0000provide an efficient 2-approximation for the STC and a 6-approximation for the\u0000STC+ minimization variants. The experiments show that, unlike standard\u0000approaches, our new highly efficient algorithms lead to consistent strong/weak\u0000labelings of the multilayer network edges.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"93 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Graph path search is a classic computer science problem that has been recently approached with Reinforcement Learning (RL) due to its potential to outperform prior methods. Existing RL techniques typically assume a global view of the network, which is not suitable for large-scale, dynamic, and privacy-sensitive settings. An area of particular interest is search in social networks due to its numerous applications. Inspired by seminal work in experimental sociology, which showed that decentralized yet efficient search is possible in social networks, we frame the problem as a collaborative task between multiple agents equipped with a limited local view of the network. We propose a multi-agent approach for graph path search that successfully leverages both homophily and structural heterogeneity. Our experiments, carried out over synthetic and real-world social networks, demonstrate that our model significantly outperforms learned and heuristic baselines. Furthermore, our results show that meaningful embeddings for graph navigation can be constructed using reward-driven learning.
{"title":"Reinforcement Learning Discovers Efficient Decentralized Graph Path Search Strategies","authors":"Alexei Pisacane, Victor-Alexandru Darvariu, Mirco Musolesi","doi":"arxiv-2409.07932","DOIUrl":"https://doi.org/arxiv-2409.07932","url":null,"abstract":"Graph path search is a classic computer science problem that has been\u0000recently approached with Reinforcement Learning (RL) due to its potential to\u0000outperform prior methods. Existing RL techniques typically assume a global view\u0000of the network, which is not suitable for large-scale, dynamic, and\u0000privacy-sensitive settings. An area of particular interest is search in social\u0000networks due to its numerous applications. Inspired by seminal work in\u0000experimental sociology, which showed that decentralized yet efficient search is\u0000possible in social networks, we frame the problem as a collaborative task\u0000between multiple agents equipped with a limited local view of the network. We\u0000propose a multi-agent approach for graph path search that successfully\u0000leverages both homophily and structural heterogeneity. Our experiments, carried\u0000out over synthetic and real-world social networks, demonstrate that our model\u0000significantly outperforms learned and heuristic baselines. Furthermore, our\u0000results show that meaningful embeddings for graph navigation can be constructed\u0000using reward-driven learning.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"18 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142214677","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In 2016, a network of social media accounts animated by Russian operatives attempted to divert political discourse within the American public around the presidential elections. This was a coordinated effort, part of a Russian-led complex information operation. Utilizing the anonymity and outreach of social media platforms Russian operatives created an online astroturf that is in direct contact with regular Americans, promoting Russian agenda and goals. The elusiveness of this type of adversarial approach rendered security agencies helpless, stressing the unique challenges this type of intervention presents. Building on existing scholarship on the functions within influence networks on social media, we suggest a new approach to map those types of operations. We argue that pretending to be legitimate social actors obliges the network to adhere to social expectations, leaving a social footprint. To test the robustness of this social footprint we train artificial intelligence to identify it and create a predictive model. We use Twitter data identified as part of the Russian influence network for training the artificial intelligence and to test the prediction. Our model attains 88% prediction accuracy for the test set. Testing our prediction on two additional models results in 90.7% and 90.5% accuracy, validating our model. The predictive and validation results suggest that building a machine learning model around social functions within the Russian influence network can be used to map its actors and functions.
{"title":"Keeping it Authentic: The Social Footprint of the Trolls Network","authors":"Ori Swed, Sachith Dassanayaka, Dimitri Volchenkov","doi":"arxiv-2409.07720","DOIUrl":"https://doi.org/arxiv-2409.07720","url":null,"abstract":"In 2016, a network of social media accounts animated by Russian operatives\u0000attempted to divert political discourse within the American public around the\u0000presidential elections. This was a coordinated effort, part of a Russian-led\u0000complex information operation. Utilizing the anonymity and outreach of social\u0000media platforms Russian operatives created an online astroturf that is in\u0000direct contact with regular Americans, promoting Russian agenda and goals. The\u0000elusiveness of this type of adversarial approach rendered security agencies\u0000helpless, stressing the unique challenges this type of intervention presents.\u0000Building on existing scholarship on the functions within influence networks on\u0000social media, we suggest a new approach to map those types of operations. We\u0000argue that pretending to be legitimate social actors obliges the network to\u0000adhere to social expectations, leaving a social footprint. To test the\u0000robustness of this social footprint we train artificial intelligence to\u0000identify it and create a predictive model. We use Twitter data identified as\u0000part of the Russian influence network for training the artificial intelligence\u0000and to test the prediction. Our model attains 88% prediction accuracy for the\u0000test set. Testing our prediction on two additional models results in 90.7% and\u000090.5% accuracy, validating our model. The predictive and validation results\u0000suggest that building a machine learning model around social functions within\u0000the Russian influence network can be used to map its actors and functions.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"319 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hypergraphs provide a robust framework for modeling complex systems with higher-order interactions. However, analyzing them in dynamic settings presents significant computational challenges. To address this, we introduce a novel method that adapts the cardinality-based gadget to convert hypergraphs into strongly connected weighted directed graphs, complemented by a symmetrized combinatorial Laplacian. We demonstrate that the harmonic mean of the conductance and edge expansion of the original hypergraph can be upper-bounded by the conductance of the transformed directed graph, effectively preserving crucial cut information. Additionally, we analyze how the resulting Laplacian relates to that derived from the star expansion. Our approach was validated through change point detection experiments on both synthetic and real datasets, showing superior performance over clique and star expansions in maintaining spectral information in dynamic settings. Finally, we applied our method to analyze a dynamic legal hypergraph constructed from extensive United States court opinion data.
{"title":"Hypergraph Change Point Detection using Adapted Cardinality-Based Gadgets: Applications in Dynamic Legal Structures","authors":"Hiroki Matsumoto, Takahiro Yoshida, Ryoma Kondo, Ryohei Hisano","doi":"arxiv-2409.08106","DOIUrl":"https://doi.org/arxiv-2409.08106","url":null,"abstract":"Hypergraphs provide a robust framework for modeling complex systems with\u0000higher-order interactions. However, analyzing them in dynamic settings presents\u0000significant computational challenges. To address this, we introduce a novel\u0000method that adapts the cardinality-based gadget to convert hypergraphs into\u0000strongly connected weighted directed graphs, complemented by a symmetrized\u0000combinatorial Laplacian. We demonstrate that the harmonic mean of the\u0000conductance and edge expansion of the original hypergraph can be upper-bounded\u0000by the conductance of the transformed directed graph, effectively preserving\u0000crucial cut information. Additionally, we analyze how the resulting Laplacian\u0000relates to that derived from the star expansion. Our approach was validated\u0000through change point detection experiments on both synthetic and real datasets,\u0000showing superior performance over clique and star expansions in maintaining\u0000spectral information in dynamic settings. Finally, we applied our method to\u0000analyze a dynamic legal hypergraph constructed from extensive United States\u0000court opinion data.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":"28 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142214674","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：