Qing Xiao, Yuhang Zheng, Xianzhe Fan, Bingbing Zhang, Zhicong Lu
Previous research pays attention to how users strategically understand and consciously interact with algorithms but mainly focuses on an individual level, making it difficult to explore how users within communities could develop a collective understanding of algorithms and organize collective algorithmic actions. Through a two-year ethnography of online fan activities, this study investigates 43 core fans who always organize large-scale fans' collective actions and their corresponding general fan groups. This study aims to reveal how these core fans mobilize millions of general fans through collective algorithmic actions. These core fans reported the rhetorical strategies used to persuade general fans, the steps taken to build a collective understanding of algorithms, and the collaborative processes that adapt collective actions across platforms and cultures. Our findings highlight the key factors that enable computer-supported collective algorithmic actions and extend collective action research into large-scale domain targeting algorithms.
{"title":"Let's Influence Algorithms Together: How Millions of Fans Build Collective Understanding of Algorithms and Organize Coordinated Algorithmic Actions","authors":"Qing Xiao, Yuhang Zheng, Xianzhe Fan, Bingbing Zhang, Zhicong Lu","doi":"arxiv-2409.10670","DOIUrl":"https://doi.org/arxiv-2409.10670","url":null,"abstract":"Previous research pays attention to how users strategically understand and\u0000consciously interact with algorithms but mainly focuses on an individual level,\u0000making it difficult to explore how users within communities could develop a\u0000collective understanding of algorithms and organize collective algorithmic\u0000actions. Through a two-year ethnography of online fan activities, this study\u0000investigates 43 core fans who always organize large-scale fans' collective\u0000actions and their corresponding general fan groups. This study aims to reveal\u0000how these core fans mobilize millions of general fans through collective\u0000algorithmic actions. These core fans reported the rhetorical strategies used to\u0000persuade general fans, the steps taken to build a collective understanding of\u0000algorithms, and the collaborative processes that adapt collective actions\u0000across platforms and cultures. Our findings highlight the key factors that\u0000enable computer-supported collective algorithmic actions and extend collective\u0000action research into large-scale domain targeting algorithms.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263436","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}
Jan von Pichowski, Christopher Blöcker, Ingo Scholtes
Graph pooling is an essential part of deep graph representation learning. We introduce MapEqPool, a principled pooling operator that takes the inherent hierarchical structure of real-world graphs into account. MapEqPool builds on the map equation, an information-theoretic objective function for community detection based on the minimum description length principle which naturally implements Occam's razor and balances between model complexity and fit. We demonstrate MapEqPool's competitive performance with an empirical comparison against various baselines across standard graph classification datasets.
{"title":"Hierarchical Graph Pooling Based on Minimum Description Length","authors":"Jan von Pichowski, Christopher Blöcker, Ingo Scholtes","doi":"arxiv-2409.10263","DOIUrl":"https://doi.org/arxiv-2409.10263","url":null,"abstract":"Graph pooling is an essential part of deep graph representation learning. We\u0000introduce MapEqPool, a principled pooling operator that takes the inherent\u0000hierarchical structure of real-world graphs into account. MapEqPool builds on\u0000the map equation, an information-theoretic objective function for community\u0000detection based on the minimum description length principle which naturally\u0000implements Occam's razor and balances between model complexity and fit. We\u0000demonstrate MapEqPool's competitive performance with an empirical comparison\u0000against various baselines across standard graph classification datasets.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263472","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}
Autoencoders based on Graph Neural Networks (GNNs) have garnered significant attention in recent years for their ability to extract informative latent representations, characterizing the structure of complex topologies, such as graphs. Despite the prevalence of Graph Autoencoders, there has been limited focus on developing and evaluating explainable neural-based graph generative models specifically designed for signed networks. To address this gap, we propose the Signed Graph Archetypal Autoencoder (SGAAE) framework. SGAAE extracts node-level representations that express node memberships over distinct extreme profiles, referred to as archetypes, within the network. This is achieved by projecting the graph onto a learned polytope, which governs its polarization. The framework employs a recently proposed likelihood for analyzing signed networks based on the Skellam distribution, combined with relational archetypal analysis and GNNs. Our experimental evaluation demonstrates the SGAAEs' capability to successfully infer node memberships over the different underlying latent structures while extracting competing communities formed through the participation of the opposing views in the network. Additionally, we introduce the 2-level network polarization problem and show how SGAAE is able to characterize such a setting. The proposed model achieves high performance in different tasks of signed link prediction across four real-world datasets, outperforming several baseline models.
{"title":"Signed Graph Autoencoder for Explainable and Polarization-Aware Network Embeddings","authors":"Nikolaos Nakis, Chrysoula Kosma, Giannis Nikolentzos, Michalis Chatzianastasis, Iakovos Evdaimon, Michalis Vazirgiannis","doi":"arxiv-2409.10452","DOIUrl":"https://doi.org/arxiv-2409.10452","url":null,"abstract":"Autoencoders based on Graph Neural Networks (GNNs) have garnered significant\u0000attention in recent years for their ability to extract informative latent\u0000representations, characterizing the structure of complex topologies, such as\u0000graphs. Despite the prevalence of Graph Autoencoders, there has been limited\u0000focus on developing and evaluating explainable neural-based graph generative\u0000models specifically designed for signed networks. To address this gap, we\u0000propose the Signed Graph Archetypal Autoencoder (SGAAE) framework. SGAAE\u0000extracts node-level representations that express node memberships over distinct\u0000extreme profiles, referred to as archetypes, within the network. This is\u0000achieved by projecting the graph onto a learned polytope, which governs its\u0000polarization. The framework employs a recently proposed likelihood for\u0000analyzing signed networks based on the Skellam distribution, combined with\u0000relational archetypal analysis and GNNs. Our experimental evaluation\u0000demonstrates the SGAAEs' capability to successfully infer node memberships over\u0000the different underlying latent structures while extracting competing\u0000communities formed through the participation of the opposing views in the\u0000network. Additionally, we introduce the 2-level network polarization problem\u0000and show how SGAAE is able to characterize such a setting. The proposed model\u0000achieves high performance in different tasks of signed link prediction across\u0000four real-world datasets, outperforming several baseline models.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263438","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}
Zhongyi Qiu, Kangyi Qiu, Hanjia Lyu, Wei Xiong, Jiebo Luo
Emojis have become an integral part of digital communication, enriching text by conveying emotions, tone, and intent. Existing emoji recommendation methods are primarily evaluated based on their ability to match the exact emoji a user chooses in the original text. However, they ignore the essence of users' behavior on social media in that each text can correspond to multiple reasonable emojis. To better assess a model's ability to align with such real-world emoji usage, we propose a new semantics preserving evaluation framework for emoji recommendation, which measures a model's ability to recommend emojis that maintain the semantic consistency with the user's text. To evaluate how well a model preserves semantics, we assess whether the predicted affective state, demographic profile, and attitudinal stance of the user remain unchanged. If these attributes are preserved, we consider the recommended emojis to have maintained the original semantics. The advanced abilities of Large Language Models (LLMs) in understanding and generating nuanced, contextually relevant output make them well-suited for handling the complexities of semantics preserving emoji recommendation. To this end, we construct a comprehensive benchmark to systematically assess the performance of six proprietary and open-source LLMs using different prompting techniques on our task. Our experiments demonstrate that GPT-4o outperforms other LLMs, achieving a semantics preservation score of 79.23%. Additionally, we conduct case studies to analyze model biases in downstream classification tasks and evaluate the diversity of the recommended emojis.
{"title":"Semantics Preserving Emoji Recommendation with Large Language Models","authors":"Zhongyi Qiu, Kangyi Qiu, Hanjia Lyu, Wei Xiong, Jiebo Luo","doi":"arxiv-2409.10760","DOIUrl":"https://doi.org/arxiv-2409.10760","url":null,"abstract":"Emojis have become an integral part of digital communication, enriching text\u0000by conveying emotions, tone, and intent. Existing emoji recommendation methods\u0000are primarily evaluated based on their ability to match the exact emoji a user\u0000chooses in the original text. However, they ignore the essence of users'\u0000behavior on social media in that each text can correspond to multiple\u0000reasonable emojis. To better assess a model's ability to align with such\u0000real-world emoji usage, we propose a new semantics preserving evaluation\u0000framework for emoji recommendation, which measures a model's ability to\u0000recommend emojis that maintain the semantic consistency with the user's text.\u0000To evaluate how well a model preserves semantics, we assess whether the\u0000predicted affective state, demographic profile, and attitudinal stance of the\u0000user remain unchanged. If these attributes are preserved, we consider the\u0000recommended emojis to have maintained the original semantics. The advanced\u0000abilities of Large Language Models (LLMs) in understanding and generating\u0000nuanced, contextually relevant output make them well-suited for handling the\u0000complexities of semantics preserving emoji recommendation. To this end, we\u0000construct a comprehensive benchmark to systematically assess the performance of\u0000six proprietary and open-source LLMs using different prompting techniques on\u0000our task. Our experiments demonstrate that GPT-4o outperforms other LLMs,\u0000achieving a semantics preservation score of 79.23%. Additionally, we conduct\u0000case studies to analyze model biases in downstream classification tasks and\u0000evaluate the diversity of the recommended emojis.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263435","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}
We propose a novel approach to assess the public's coping behavior during the COVID-19 outbreak by examining the emotions. Specifically, we explore (1) changes in the public's emotions with the COVID-19 crisis progression and (2) the impacts of the public's emotions on their information-seeking, information-sharing behaviors, and compliance with stay-at-home policies. We base the study on the appraisal tendency framework, detect the public's emotions by fine-tuning a pre-trained RoBERTa model, and cross-analyze third-party behavioral data. We demonstrate the feasibility and reliability of our proposed approach in providing a large-scale examination of the publi's emotions and coping behaviors in a real-world crisis: COVID-19. The approach complements prior crisis communication research, mainly based on self-reported, small-scale experiments and survey data. Our results show that anger and fear are more prominent than other emotions experienced by the public at the pandemic's outbreak stage. Results also show that the extent of low certainty and passive emotions (e.g., sadness, fear) was related to increased information-seeking and information-sharing behaviors. Additionally, high-certainty (e.g., anger) and low-certainty (e.g., sadness, fear) emotions during the outbreak correlated to the public's compliance with stay-at-home orders.
我们提出了一种新方法，通过研究情绪来评估公众在 COVID-19 爆发期间的应对行为。具体来说，我们探讨了（1）公众情绪随着 COVID-19 危机进展的变化；（2）公众情绪对其信息搜寻、信息共享行为和遵守留守政策的影响。我们的研究基于评价倾向框架，通过微调预先训练好的 RoBERTa 模型来检测公众情绪，并交叉分析第三方行为数据。我们证明了所提出方法的可行性和可靠性，可以对真实世界危机中的公众情绪和应对行为进行大规模检测：COVID-19。该方法补充了之前的危机传播研究，这些研究主要基于自我报告、小规模实验和调查数据。我们的研究结果表明，在疫情爆发阶段，公众的愤怒和恐惧情绪比其他情绪更为突出。结果还显示，低确定性和被动情绪（如悲伤、恐惧）的程度与信息搜寻和信息分享行为的增加有关。此外，疫情爆发期间的高确定性情绪（如愤怒）和低确定性情绪（如悲伤、恐惧）与公众遵守留在家中的命令有关。
{"title":"Impact Of Emotions on Information Seeking And Sharing Behaviors During Pandemic","authors":"Smitha Muthya Sudheendra, Hao Xu, Jisu Huh, Jaideep Srivastava","doi":"arxiv-2409.10754","DOIUrl":"https://doi.org/arxiv-2409.10754","url":null,"abstract":"We propose a novel approach to assess the public's coping behavior during the\u0000COVID-19 outbreak by examining the emotions. Specifically, we explore (1)\u0000changes in the public's emotions with the COVID-19 crisis progression and (2)\u0000the impacts of the public's emotions on their information-seeking,\u0000information-sharing behaviors, and compliance with stay-at-home policies. We\u0000base the study on the appraisal tendency framework, detect the public's\u0000emotions by fine-tuning a pre-trained RoBERTa model, and cross-analyze\u0000third-party behavioral data. We demonstrate the feasibility and reliability of\u0000our proposed approach in providing a large-scale examination of the publi's\u0000emotions and coping behaviors in a real-world crisis: COVID-19. The approach\u0000complements prior crisis communication research, mainly based on self-reported,\u0000small-scale experiments and survey data. Our results show that anger and fear\u0000are more prominent than other emotions experienced by the public at the\u0000pandemic's outbreak stage. Results also show that the extent of low certainty\u0000and passive emotions (e.g., sadness, fear) was related to increased\u0000information-seeking and information-sharing behaviors. Additionally,\u0000high-certainty (e.g., anger) and low-certainty (e.g., sadness, fear) emotions\u0000during the outbreak correlated to the public's compliance with stay-at-home\u0000orders.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142269353","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}
The ability of Graph Neural Networks (GNNs) to capture long-range and global topology information is limited by the scope of conventional graph Laplacian, leading to unsatisfactory performance on some datasets, particularly on heterophilic graphs. To address this limitation, we propose a new class of parameterized Laplacian matrices, which provably offers more flexibility in controlling the diffusion distance between nodes than the conventional graph Laplacian, allowing long-range information to be adaptively captured through diffusion on graph. Specifically, we first prove that the diffusion distance and spectral distance on graph have an order-preserving relationship. With this result, we demonstrate that the parameterized Laplacian can accelerate the diffusion of long-range information, and the parameters in the Laplacian enable flexibility of the diffusion scopes. Based on the theoretical results, we propose topology-guided rewiring mechanism to capture helpful long-range neighborhood information for heterophilic graphs. With this mechanism and the new Laplacian, we propose two GNNs with flexible diffusion scopes: namely the Parameterized Diffusion based Graph Convolutional Networks (PD-GCN) and Graph Attention Networks (PD-GAT). Synthetic experiments reveal the high correlations between the parameters of the new Laplacian and the performance of parameterized GNNs under various graph homophily levels, which verifies that our new proposed GNNs indeed have the ability to adjust the parameters to adaptively capture the global information for different levels of heterophilic graphs. They also outperform the state-of-the-art (SOTA) models on 6 out of 7 real-world benchmark datasets, which further confirms their superiority.
{"title":"Flexible Diffusion Scopes with Parameterized Laplacian for Heterophilic Graph Learning","authors":"Qincheng Lu, Jiaqi Zhu, Sitao Luan, Xiao-Wen Chang","doi":"arxiv-2409.09888","DOIUrl":"https://doi.org/arxiv-2409.09888","url":null,"abstract":"The ability of Graph Neural Networks (GNNs) to capture long-range and global\u0000topology information is limited by the scope of conventional graph Laplacian,\u0000leading to unsatisfactory performance on some datasets, particularly on\u0000heterophilic graphs. To address this limitation, we propose a new class of\u0000parameterized Laplacian matrices, which provably offers more flexibility in\u0000controlling the diffusion distance between nodes than the conventional graph\u0000Laplacian, allowing long-range information to be adaptively captured through\u0000diffusion on graph. Specifically, we first prove that the diffusion distance\u0000and spectral distance on graph have an order-preserving relationship. With this\u0000result, we demonstrate that the parameterized Laplacian can accelerate the\u0000diffusion of long-range information, and the parameters in the Laplacian enable\u0000flexibility of the diffusion scopes. Based on the theoretical results, we\u0000propose topology-guided rewiring mechanism to capture helpful long-range\u0000neighborhood information for heterophilic graphs. With this mechanism and the\u0000new Laplacian, we propose two GNNs with flexible diffusion scopes: namely the\u0000Parameterized Diffusion based Graph Convolutional Networks (PD-GCN) and Graph\u0000Attention Networks (PD-GAT). Synthetic experiments reveal the high correlations\u0000between the parameters of the new Laplacian and the performance of\u0000parameterized GNNs under various graph homophily levels, which verifies that\u0000our new proposed GNNs indeed have the ability to adjust the parameters to\u0000adaptively capture the global information for different levels of heterophilic\u0000graphs. They also outperform the state-of-the-art (SOTA) models on 6 out of 7\u0000real-world benchmark datasets, which further confirms their superiority.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263473","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}
Financial fraud refers to the act of obtaining financial benefits through dishonest means. Such behavior not only disrupts the order of the financial market but also harms economic and social development and breeds other illegal and criminal activities. With the popularization of the internet and online payment methods, many fraudulent activities and money laundering behaviors in life have shifted from offline to online, posing a great challenge to regulatory authorities. How to efficiently detect these financial fraud activities has become an urgent issue that needs to be resolved. Graph neural networks are a type of deep learning model that can utilize the interactive relationships within graph structures, and they have been widely applied in the field of fraud detection. However, there are still some issues. First, fraudulent activities only account for a very small part of transaction transfers, leading to an inevitable problem of label imbalance in fraud detection. At the same time, fraudsters often disguise their behavior, which can have a negative impact on the final prediction results. In addition, existing research has overlooked the importance of balancing neighbor information and central node information. For example, when the central node has too many neighbors, the features of the central node itself are often neglected. Finally, fraud activities and patterns are constantly changing over time, so considering the dynamic evolution of graph edge relationships is also very important.
{"title":"Dynamic Fraud Detection: Integrating Reinforcement Learning into Graph Neural Networks","authors":"Yuxin Dong, Jianhua Yao, Jiajing Wang, Yingbin Liang, Shuhan Liao, Minheng Xiao","doi":"arxiv-2409.09892","DOIUrl":"https://doi.org/arxiv-2409.09892","url":null,"abstract":"Financial fraud refers to the act of obtaining financial benefits through\u0000dishonest means. Such behavior not only disrupts the order of the financial\u0000market but also harms economic and social development and breeds other illegal\u0000and criminal activities. With the popularization of the internet and online\u0000payment methods, many fraudulent activities and money laundering behaviors in\u0000life have shifted from offline to online, posing a great challenge to\u0000regulatory authorities. How to efficiently detect these financial fraud\u0000activities has become an urgent issue that needs to be resolved. Graph neural\u0000networks are a type of deep learning model that can utilize the interactive\u0000relationships within graph structures, and they have been widely applied in the\u0000field of fraud detection. However, there are still some issues. First,\u0000fraudulent activities only account for a very small part of transaction\u0000transfers, leading to an inevitable problem of label imbalance in fraud\u0000detection. At the same time, fraudsters often disguise their behavior, which\u0000can have a negative impact on the final prediction results. In addition,\u0000existing research has overlooked the importance of balancing neighbor\u0000information and central node information. For example, when the central node\u0000has too many neighbors, the features of the central node itself are often\u0000neglected. Finally, fraud activities and patterns are constantly changing over\u0000time, so considering the dynamic evolution of graph edge relationships is also\u0000very important.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263440","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}
Priya Ronald D'Costa, Evan Rowbotham, Xinlan Emily Hu
When conflicts escalate, is it due to what is said or how it is said? In the conflict literature, two theoretical approaches take opposing views: one focuses on the content of the disagreement, while the other focuses on how it is expressed. This paper aims to integrate these two perspectives through a computational analysis of 191 communication features -- 128 related to expression and 63 to content. We analyze 1,200 GPT-4 simulated conversations and 12,630 real-world discussions from Reddit. We find that expression features more reliably predict destructive conflict outcomes across both settings, although the most important features differ. In the Reddit data, conversational dynamics such as turn-taking and conversational equality are highly predictive, but they are not predictive in simulated conversations. These results may suggest a possible limitation in simulating social interactions with language models, and we discuss the implications for our findings on building social computing systems.
{"title":"What you say or how you say it? Predicting Conflict Outcomes in Real and LLM-Generated Conversations","authors":"Priya Ronald D'Costa, Evan Rowbotham, Xinlan Emily Hu","doi":"arxiv-2409.09338","DOIUrl":"https://doi.org/arxiv-2409.09338","url":null,"abstract":"When conflicts escalate, is it due to what is said or how it is said? In the\u0000conflict literature, two theoretical approaches take opposing views: one\u0000focuses on the content of the disagreement, while the other focuses on how it\u0000is expressed. This paper aims to integrate these two perspectives through a\u0000computational analysis of 191 communication features -- 128 related to\u0000expression and 63 to content. We analyze 1,200 GPT-4 simulated conversations\u0000and 12,630 real-world discussions from Reddit. We find that expression features\u0000more reliably predict destructive conflict outcomes across both settings,\u0000although the most important features differ. In the Reddit data, conversational\u0000dynamics such as turn-taking and conversational equality are highly predictive,\u0000but they are not predictive in simulated conversations. These results may\u0000suggest a possible limitation in simulating social interactions with language\u0000models, and we discuss the implications for our findings on building social\u0000computing systems.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142269902","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}
Lingyao Li, Songhua Hu, Yinpei Dai, Min Deng, Parisa Momeni, Gabriel Laverghetta, Lizhou Fan, Zihui Ma, Xi Wang, Siyuan Ma, Jay Ligatti, Libby Hemphill
As urban populations grow, the need for accessible urban design has become urgent. Traditional survey methods for assessing public perceptions of accessibility are often limited in scope. Crowdsourcing via online reviews offers a valuable alternative to understanding public perceptions, and advancements in large language models can facilitate their use. This study uses Google Maps reviews across the United States and fine-tunes Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility. At the POI level, most categories -- restaurants, retail, hotels, and healthcare -- show negative sentiments. Socio-spatial analysis reveals that areas with higher proportions of white residents and greater socioeconomic status report more positive sentiment, while areas with more elderly, highly-educated residents exhibit more negative sentiment. Interestingly, no clear link is found between the presence of disabilities and public sentiments. Overall, this study highlights the potential of crowdsourcing for identifying accessibility challenges and providing insights for urban planners.
{"title":"Toward satisfactory public accessibility: A crowdsourcing approach through online reviews to inclusive urban design","authors":"Lingyao Li, Songhua Hu, Yinpei Dai, Min Deng, Parisa Momeni, Gabriel Laverghetta, Lizhou Fan, Zihui Ma, Xi Wang, Siyuan Ma, Jay Ligatti, Libby Hemphill","doi":"arxiv-2409.08459","DOIUrl":"https://doi.org/arxiv-2409.08459","url":null,"abstract":"As urban populations grow, the need for accessible urban design has become\u0000urgent. Traditional survey methods for assessing public perceptions of\u0000accessibility are often limited in scope. Crowdsourcing via online reviews\u0000offers a valuable alternative to understanding public perceptions, and\u0000advancements in large language models can facilitate their use. This study uses\u0000Google Maps reviews across the United States and fine-tunes Llama 3 model with\u0000the Low-Rank Adaptation technique to analyze public sentiment on accessibility.\u0000At the POI level, most categories -- restaurants, retail, hotels, and\u0000healthcare -- show negative sentiments. Socio-spatial analysis reveals that\u0000areas with higher proportions of white residents and greater socioeconomic\u0000status report more positive sentiment, while areas with more elderly,\u0000highly-educated residents exhibit more negative sentiment. Interestingly, no\u0000clear link is found between the presence of disabilities and public sentiments.\u0000Overall, this study highlights the potential of crowdsourcing for identifying\u0000accessibility challenges and providing insights for urban planners.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263481","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 an Online Social Network (OSN), users can create a unique public persona by crafting a user identity that may encompass profile details, content, and network-related information. As a result, a relevant task of interest is related to the ability to link identities across different OSNs. Linking users across social networks can have multiple implications in several contexts both at the individual level and at the group level. At the individual level, the main interest in linking the same identity across social networks is to enable a better knowledge of each user. At the group level, linking user identities through different OSNs helps in predicting user behaviors, network dynamics, information diffusion, and migration phenomena across social media. The process of tying together user accounts on different OSNs is challenging and has attracted more and more research attention in the last fifteen years. The purpose of this work is to provide a comprehensive review of recent studies (from 2016 to the present) on User Identity Linkage (UIL) methods across online social networks. This review aims to offer guidance for other researchers in the field by outlining the main problem formulations, the different feature extraction strategies, algorithms, machine learning models, datasets, and evaluation metrics proposed by researchers working in this area. The proposed overview takes a pragmatic perspective to highlight the concrete possibilities for accomplishing this task depending on the type of available data.
在在线社交网络（OSN）中，用户可以创建一个独一无二的公共个人信息，精心制作用户身份，其中可能包括个人资料、内容和网络相关信息。因此，人们感兴趣的一项相关任务与在不同的 OSN 之间链接身份的能力有关。跨社交网络链接用户在个人和群体两个层面都会产生多重影响。在个人层面，跨社交网络链接同一身份的主要目的是为了更好地了解每个用户。在群体层面，通过不同的 OSNs 链接用户身份有助于预测用户行为、网络动态、信息扩散和社交媒体间的迁移现象。将不同 OSN 上的用户账户绑定在一起的过程极具挑战性，在过去 15 年里吸引了越来越多的研究关注。这项工作的目的是全面回顾近期（2016 年至今）关于跨网络社交网络用户身份关联（UIL）方法的研究。本综述旨在通过概述该领域研究人员提出的主要问题表述、不同的特征提取策略、算法、机器学习模型、数据集和评估指标，为该领域的其他研究人员提供指导。本综述从务实的角度出发，强调了根据可用数据类型完成这项任务的具体可能性。
{"title":"User Identity Linkage on Social Networks: A Review of Modern Techniques and Applications","authors":"Caterina Senette, Marco Siino, Maurizio Tesconi","doi":"arxiv-2409.08966","DOIUrl":"https://doi.org/arxiv-2409.08966","url":null,"abstract":"In an Online Social Network (OSN), users can create a unique public persona\u0000by crafting a user identity that may encompass profile details, content, and\u0000network-related information. As a result, a relevant task of interest is\u0000related to the ability to link identities across different OSNs. Linking users\u0000across social networks can have multiple implications in several contexts both\u0000at the individual level and at the group level. At the individual level, the\u0000main interest in linking the same identity across social networks is to enable\u0000a better knowledge of each user. At the group level, linking user identities\u0000through different OSNs helps in predicting user behaviors, network dynamics,\u0000information diffusion, and migration phenomena across social media. The process\u0000of tying together user accounts on different OSNs is challenging and has\u0000attracted more and more research attention in the last fifteen years. The\u0000purpose of this work is to provide a comprehensive review of recent studies\u0000(from 2016 to the present) on User Identity Linkage (UIL) methods across online\u0000social networks. This review aims to offer guidance for other researchers in\u0000the field by outlining the main problem formulations, the different feature\u0000extraction strategies, algorithms, machine learning models, datasets, and\u0000evaluation metrics proposed by researchers working in this area. The proposed\u0000overview takes a pragmatic perspective to highlight the concrete possibilities\u0000for accomplishing this task depending on the type of available data.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142263475","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：