abstract: Sarcasm is a nuanced form of language where usually, the speaker explicitly states the opposite of what is implied. Imbued with intentional ambiguity and subtlety, detecting sarcasm is a difficult task, even for humans. Current works approach this challenging problem primarily from a linguistic perspective, focusing on the lexical and syntactic aspects of sarcasm. In this thesis, I explore the possibility of using behavior traits intrinsic to users of sarcasm to detect sarcastic tweets. First, I theorize the core forms of sarcasm using findings from the psychological and behavioral sciences, and some observations on Twitter users. Then, I develop computational features to model the manifestations of these forms of sarcasm using the user's profile information and tweets. Finally, I combine these features to train a supervised learning model to detect sarcastic tweets. I perform experiments to extensively evaluate the proposed behavior modeling approach and compare with the state-of-the-art. / Dissertation/Thesis / Masters Thesis Computer Science 2014

Metódos analíticos de investigação são usualmente ineficazes para sistemas computacionais sociais já que apenas algumas iterações do sistema já são suficientes para que o sistema se torne imprevisível. Portanto, uma das principais questões na Computação Social é o desenvolvimento de modelos sociais passíveis de investigação. Assim é possível que se compreenda o relacionamento complexo entre os componentes de sistemas sociais computacionais e o resultado. Este aspectos incluem a modelagem, a estrutura de comunicação e características individuais do agentes envolvidos na resolução dos problemas. do processo social. Esta tese explora sistemas computacionais de resolução de problemas com foco em sistemas artificiais e heterogêneos. Nela é feita uma compilação extensiva da literatura relacionada em sistemas complexos onde as contribuições do candidato são expostas dentro de contextos específicos da área. Entre elas está o estudo de modelos abstratos e gerais de resolução social de problemas, a investigação do impacto da centralidade no resultado individual e coletivo, a análise experimental de modelos heterogêneos de resolução social de problemas. Quando integradas, estas contribuições reforçam o entendimento sobre a importância da rede e das estruturas de comunicação, a composição estratégica do sistema, a estrutura do problema e possíveis padrões gerais na resolução social de problemas. / This thesis reviews and investigates social problem-solving with a particular focus on artificial and heterogeneous systems. More specifically, we not only compile and comprehensively examine recent research results, but also discuss future directions in the study of such heterogeneous complex systems. Given their complex nature, such systems often defy analyses. Even computationally simple models can behave unpredictably after a few iterations. Therefore, one central issue in Social Computing is to devise models of social interaction that are amenable to investigation. This way, one can understand the complex relationships among the components and the outcome of the social process. This thesis surveys scientific inquiries concerned with fundamental aspects in social problemsolving systems and their impact in ability and performance of such systems. These aspects include modeling, communication structure and individual problem-solver traits. This thesis also reports the student endeavour during the period of research and summarizes several already published contributions. Among them there is (i) the study of general frameworks for the study of social problem-solving, (ii) the investigation of the role of centrality in individual and collective outcomes, and (iii) the exploration of heterogeneous models of social problem-solving. These three points, in an integrated perspective underpin the understanding of network and communication structures, adjust the strategic systems’ composition, and exploit problems’ structures and patterns in social problemsolving systems.

Nossa incapacidade em compreender todos os fatores responsáveis por fenômenos naturais faz com que tenhamos que recorrer a simplificações na representação e na explicação destes. Por sua vez, a forma com que representamos e pensamos a respeito destes fenômenos é influenciada por fatores de natureza interna, como o nosso estado psicológico, ou então de natureza externa, como o ambiente social. Dentre os fatores externos, o ambiente social, ou contexto social, é um dos que tem maior influência na forma que pensamos e agimos. Quando estamos em grupo, mudamos a todo instante a forma com que resolvemos problemas em resposta ao contexto que nos cerca. Entretanto, esta característica até então foi pouco explorada em modelos computacionais de resolução coletiva de problemas. Este trabalho investiga o impacto do contexto social na resolução coletiva de problemas. Nós apresentaremos evidências de que o contexto social tem um papel importante na forma com que o grupo e o indivíduos se comportam. Mais precisamente, nós mostraremos que a centralidade de um indivíduo na rede social nem sempre é um bom preditor de sua contribuição quando o mesmo pode adaptar sua estratégia de busca em resposta ao contexto. Além disso, mostraremos que a adaptação ao contexto social por parte dos indivíduos pode melhorar o desempenho coletivo, facilitando a convergência para soluções boas; e que a diversidade de estratégias de resolução do problema não leva necessariamente a uma diversidade de soluções na população; e que, mesmo que o contexto social seja percebido da mesma forma pelos indivíduos, a forma com que eles reagem pode levar a diferentes resultados. Todos estes resultados suportam a ideia de que o contexto social deve ser considerado em experimentos com resolução social de problemas. Por fim, concluímos o trabalho discutindo o impactso do mesmo e apontando novos problemas a serem investigados. / Our inability to perceive and understand all the factors that account for real-world phenomena forces us to rely on clues when reasoning and making decisions about the world. Clues can be internal such as our psychological state and our motivations; or external, such as the resources available, the physical environment, the social environment, etc. The social environment, or social context, encompasses the set of relationships and cultural settings by which we interact and function in a society. Much of our thinking is influenced by the social environment and we constantly change the way we solve problems in response to our social environment. Nevertheless, this human trait has not been thoughtfully investigated by current computational models of human social problem-solving, for these models have lacked the heterogeneity and self-adaptive behavior observed in humans. In this work, we address this issue by investigating the impact of social context in social problem solving by means of extensive numerical simulations using a modified social model. We show evidences that social context plays a key role in how the system behaves and performs. More precisely, we show that the centrality of an agent in the network is an unreliable predictor the agent’s contribution when this agent can change its problem-solving strategy according to social context. Another finding is that social context information can be used to improve the convergence speed of the group to good solutions and that diversity in search strategies does not necessarily translates into diversity in solutions. We also determine that even if nodes perceive social context in same way, the way they react to it may lead to different outcomes along the search process. Together, these results contribute to the understanding that social context does indeed impact in social problem-solving. We conclude discussing the overall impact of this work and pointing future directions.

Nossa incapacidade em compreender todos os fatores responsáveis por fenômenos naturais faz com que tenhamos que recorrer a simplificações na representação e na explicação destes. Por sua vez, a forma com que representamos e pensamos a respeito destes fenômenos é influenciada por fatores de natureza interna, como o nosso estado psicológico, ou então de natureza externa, como o ambiente social. Dentre os fatores externos, o ambiente social, ou contexto social, é um dos que tem maior influência na forma que pensamos e agimos. Quando estamos em grupo, mudamos a todo instante a forma com que resolvemos problemas em resposta ao contexto que nos cerca. Entretanto, esta característica até então foi pouco explorada em modelos computacionais de resolução coletiva de problemas. Este trabalho investiga o impacto do contexto social na resolução coletiva de problemas. Nós apresentaremos evidências de que o contexto social tem um papel importante na forma com que o grupo e o indivíduos se comportam. Mais precisamente, nós mostraremos que a centralidade de um indivíduo na rede social nem sempre é um bom preditor de sua contribuição quando o mesmo pode adaptar sua estratégia de busca em resposta ao contexto. Além disso, mostraremos que a adaptação ao contexto social por parte dos indivíduos pode melhorar o desempenho coletivo, facilitando a convergência para soluções boas; e que a diversidade de estratégias de resolução do problema não leva necessariamente a uma diversidade de soluções na população; e que, mesmo que o contexto social seja percebido da mesma forma pelos indivíduos, a forma com que eles reagem pode levar a diferentes resultados. Todos estes resultados suportam a ideia de que o contexto social deve ser considerado em experimentos com resolução social de problemas. Por fim, concluímos o trabalho discutindo o impactso do mesmo e apontando novos problemas a serem investigados. / Our inability to perceive and understand all the factors that account for real-world phenomena forces us to rely on clues when reasoning and making decisions about the world. Clues can be internal such as our psychological state and our motivations; or external, such as the resources available, the physical environment, the social environment, etc. The social environment, or social context, encompasses the set of relationships and cultural settings by which we interact and function in a society. Much of our thinking is influenced by the social environment and we constantly change the way we solve problems in response to our social environment. Nevertheless, this human trait has not been thoughtfully investigated by current computational models of human social problem-solving, for these models have lacked the heterogeneity and self-adaptive behavior observed in humans. In this work, we address this issue by investigating the impact of social context in social problem solving by means of extensive numerical simulations using a modified social model. We show evidences that social context plays a key role in how the system behaves and performs. More precisely, we show that the centrality of an agent in the network is an unreliable predictor the agent’s contribution when this agent can change its problem-solving strategy according to social context. Another finding is that social context information can be used to improve the convergence speed of the group to good solutions and that diversity in search strategies does not necessarily translates into diversity in solutions. We also determine that even if nodes perceive social context in same way, the way they react to it may lead to different outcomes along the search process. Together, these results contribute to the understanding that social context does indeed impact in social problem-solving. We conclude discussing the overall impact of this work and pointing future directions.

abstract: Social media has become an important means of user-centered information sharing and communications in a gamut of domains, including news consumption, entertainment, marketing, public relations, and many more. The low cost, easy access, and rapid dissemination of information on social media draws a large audience but also exacerbate the wide propagation of disinformation including fake news, i.e., news with intentionally false information. Disinformation on social media is growing fast in volume and can have detrimental societal effects. Despite the importance of this problem, our understanding of disinformation in social media is still limited. Recent advancements of computational approaches on detecting disinformation and fake news have shown some early promising results. Novel challenges are still abundant due to its complexity, diversity, dynamics, multi-modality, and costs of fact-checking or annotation. Social media data opens the door to interdisciplinary research and allows one to collectively study large-scale human behaviors otherwise impossible. For example, user engagements over information such as news articles, including posting about, commenting on, or recommending the news on social media, contain abundant rich information. Since social media data is big, incomplete, noisy, unstructured, with abundant social relations, solely relying on user engagements can be sensitive to noisy user feedback. To alleviate the problem of limited labeled data, it is important to combine contents and this new (but weak) type of information as supervision signals, i.e., weak social supervision, to advance fake news detection. The goal of this dissertation is to understand disinformation by proposing and exploiting weak social supervision for learning with little labeled data and effectively detect disinformation via innovative research and novel computational methods. In particular, I investigate learning with weak social supervision for understanding disinformation with the following computational tasks: bringing the heterogeneous social context as auxiliary information for effective fake news detection; discovering explanations of fake news from social media for explainable fake news detection; modeling multi-source of weak social supervision for early fake news detection; and transferring knowledge across domains with adversarial machine learning for cross-domain fake news detection. The findings of the dissertation significantly expand the boundaries of disinformation research and establish a novel paradigm of learning with weak social supervision that has important implications in broad applications in social media. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2020

Virtual influencers (VIs), computer-generated characters, can act as a medium to influence followers on social media platforms. This is a relatively novel phenomenon, with popular VIs having over six million followers. This study investigates VIs from the perspective of human-computer interaction. Nine participants from four countries were interviewed about their interactions with a variety of fictional VIs in different settings. The study uncovers new user perceptions and attitudes towards VIs, including pessimism towards VIs, loss of trust and likeability towards virtual characters, and the perception of VIs as being run by artificial intelligence (AI). It was found that content relating to fitness, personal tips, and social activism demands more transparency about the authority and agency of the VI from the users. Finally, this study discusses the ethical implications of VIs, highlights an avenue of research in the AI domain, and makes suggestions for human-computer interaction (HCI) practitioners curating social media experiences. / Virtuella influencers (VI), datorgenererade karaktärer, kan fungera som ett medium för att påverka följare på sociala medieplattformar. Detta är ett relativt nytt fenomen, där populära VI har över sex miljoner följare. Denna studie undersöker fenomenet VI ur människa-datorinteraktions perspektiv. Nio deltagare från fyra länder intervjuades om deras interaktioner med olika fiktiva VI i olika miljöer. Studien visar på nya uppfattningar och attityder från användare gentemot VI, inklusive pessimism gentemot VI, förlust av förtroende och gillande gentemot virtuella karaktärer, samt uppfattningen om att VI drivs av artificiell intelligens (AI). Det visade sig att användarna kräver att innehåll relaterat till träning, personliga tips och social aktivism kräver mer öppenhet om VI:s befogenhet och agenda. Slutligen diskuterar denna studie de etiska konsekvenserna av VI, belyser ett forskningsområde inom AI och ger förslag till utövare inom människa-datorinteraktion (HCI) som kuraterar sociala medieupplevelser

No description available.

The rapid developments Social Network Service (SNS) have offered opportunities to re-visit many seminal theoretical assumptions of technology usage within socio-technical environment. Online social network is a rapidly growing field that imposes new questions to the existing IS research paradigm. It is argued that information systems research must necessarily evolve in response to the emerging trends (Lyytinen and King, 2004). Specifically, one stream of research has been heavily lacking is SNS usage prediction (Hargittai, 2007). In SNS, the form of social network is realized by computing networks where the individual assumes an identity of an “avatar”. People are merging their activities of work and living thus blurring the borders among their social contexts (Beck and Wade, 2006). Such new trends have become more sophisticated due to the increasingly robust data network capacity and pervasive availability of communication technology. At present, research in SNS is still in its early stage; hence the need to develop knowledge of virtual world dynamics has become impending. SNS essentially provides various service channels to facilitate social network interactions. These channels are highly correlated with their respective service contexts, among which differences are obvious and important. Cultural factors have been crucial for context oriented studies in both IS and sociology fields. For instance, the global nature of the Internet “raises questions about the robustness of trust effects across cultures” (Jarvenpaa et al., 1999). In SNS, not only trust but also privacy has become a tremendous caveat for service providers. Investigating the combination effects of privacy and trust in a cross-cultural study may lead to important theoretical discoveries and meaningful managerial implications. This study contributes to knowledge by empirically testing established theoretical models of IS acceptance, trust, social and cultural research. Both qualitative and quantitative methods are used in order to present a comprehensive analysis of SNS perception and use in different cultural settings. Particularly, the study finds critical differences exist within the process of trust formulation between American and Chinese SNS users.

Social media platforms have become a popular and inexpensive way for people to communicate with millions of others. However, this increased usage has also led to an increase in risks associated with it, such as cyberbullying, trolling, misinformation, and privacy abuse. Previous research in this field has mainly focused on isolated aspects of online risk detection, which can limit the effectiveness of these studies. To address these issues, this dissertation presents a more holistic approach to detecting and reducing harmful and abusive behavior online. To gain an initial understanding of the problem, we first present a mixed-method study of messages and media files shared in private conversations by youth to understand the risky communication experienced by them. We use these findings to determine which features can automatically detect unsafe private conversations and whether social media platforms can implement such a system given the recent move towards end-to-end encryption. We present an ensemble machine learning classifier to detect risks in private messages and how to incorporate child safety by design. In the second part of this thesis, we will explore ways to stay ahead of hate and toxicity, given the changing online behaviors. Toxic language changes over time, with aggressors inventing new insults and abusive terms that frequently target certain vulnerable communities, including women and minorities. We develop automated systems that, given an initial lexicon of toxic speech, can learn new and emerging toxic words by observing conversations on social networks. Lastly, we examine the cross-platform implications of employing risk detection systems online. Most of the research focuses only on malicious activity that occurs on one platform, which does not allow us to get a full picture of the problem. Users are obviously not bound to a single platform but can migrate to other online services for example, anecdotal evidence shows that once hateful users are banned from Twitter, they often move to Gab, an alternative social network with an open lack of moderation marketed as protection of "free speech". Consequently, we argue that moderation efforts should extend beyond safeguarding users on individual platforms and account for the potential adverse consequences of banning users from prominent platforms.

Les organisations adoptent de plus en plus les Systèmes (PAIS) pour gérer leurs processus métiers basés sur les services en utilisant les modèles de processus appelés «modèles de processus métiers». Motivés par l’adaptation aux exigences commerciales et par la réduction des coûts de maintenance, les organisations externalisent leurs processus dans le Cloud Computing. Selon l'Institut NIST, Cloud Computing est un modèle qui permet aux fournisseurs de partager leurs ressources et aux utilisateurs d’y accéder de manière pratique et à la demande. Dans un tel environnement multi-tenant, l'utilisation de modèles de processus configurables permet aux fournisseurs de processus Cloud de fournir un processus personnalisable qui peut être configuré par différents tenants en fonction de leurs besoins.Un processus métier peut être spécifié par plusieurs perspectives tel que la perspective de flux de contrôle, la perspective des ressources, etc. Plusieurs approches ont été proposées au niveau des premières perspectives, notamment le flux de contrôle. Cependant, la perspective ressource, qui est d'une importance égale, était négligée et pas explicitement définie. D’un côté, la gestion de la perspective ressource spécifiquement l’allocation des ressources Cloud est un thème d’actualité qui implique plusieurs recherches. La modélisation et la configuration des ressources sont une tâche sensible nécessitant un travail intensif. Malgré l’existence de différentes approches, elles traitent principalement les ressources humaines plutôt que des ressources Cloud. D’un autre côté, malgré le fait que le concept des modèles de processus configurables est très complémentaire au Cloud, la manière dont comment les ressources sont configurées et intégrées est à peine manipulée. Les approches proposées travaillant sur l’extension de la configuration de ressources, ne couvrent pas les propriétés Cloud notamment l’élasticité et le partage.Pour répondre à ces lacunes, nous proposons une approche pour supporter la modélisation et la configuration de l’allocation des ressources Cloud dans les modèles de processus configurables. Nous visons à (1) définir une description unifiée et formelle pour la perspective ressource, (2) assurer une allocation de ressource correcte, sans conflits et optimisée, (3) Aider les fournisseurs de processus à concevoir leur allocation de ressources configurable de manière fine afin d'éviter des résultats complexes et importants, et (4) Optimiser la sélection des ressources Cloud par rapport aux exigences liées aux propriétés Cloud (élasticité et partage) et propriétés QoS.Pour ce faire, nous proposons d'abord un cadre sémantique pour une description de ressources sémantiquement enrichies dans les processus métiers visant à formaliser les ressources Cloud consommées à l'aide d'une base de connaissances partagée. Ensuite, nous nous basons sur les processus métiers sociales pour fournir des stratégies afin d'assurer une allocation de ressources contrôlée sans conflits en termes de ressources. Par la suite, nous proposons une nouvelle approche qui étend les modèles de processus configurables pour permettre une allocation de ressources Cloud configurable. Notre objectif est de déplacer l'allocation de ressources Cloud du côté des tenants vers le côté du fournisseur de processus Cloud pour une gestion centralisée des ressources. Après, nous proposons des approches génétiques qui visent à choisir une configuration optimale des ressources d'une manière efficace sur le plan énergétique en améliorant les propriétés QoS.Afin de montrer l'efficacité de nos propositions, nous avons développé concrètement (1) une série de preuves de concepts, en tant que partie de validation, pour aider à concevoir des modèles de processus et remplir une base de connaissances de modèles de processus hétérogènes avec des ressources Cloud et (2) ont effectué des expériences sur des modèles de processus réels à partir de grands ensembles de données / Organizations are recently more and more adopting Process-Aware Information Systems (PAIS) for managing their service-based processes using process models referred to as business process models. Motivated by adapting to the rapid changing business requirements and reducing maintenance costs, organizations are outsourcing their processes in an important infrastructure which is Cloud Computing. According to the NIST Institute, Cloud Computing is a model that enables providers sharing their computing resources (e.g., networks, applications, and storage) and users accessing them in convenient and on-demand way with a minimal management effort. In such a multi-tenant environment, using configurable process models allows a Cloud process provider to deliver a customizable process that can be configured by different tenants according to their needs.A business process could be specified from various perspectives such as the control-flow perspective, the organizational perspective, the resource perspective, etc. Several approaches have been correctly proposed at the level of the first perspectives, in particular the control-flow, i.e., the temporal ordering of the process activities. Nevertheless, the resource perspective, which is of equal importance, has been neglected and poorly operated. The management of the resource perspective especially the Cloud resource allocation in business processes is a current interesting topic that increasingly involves many researches in both academics and industry. The design and configuration of resources are undoubtedly sensitive and labor-intensive task. On the one hand, the resource perspective in process models is not explicitly defined. Although many proposals exist in the literature, they all targeted human resources rather than Cloud resources. On the other hand, despite of the fact that the concept of configurable process models is highly complementary to Cloud Computing, the way in how resources can be configured and integrated is hardly handled. The few proposals, which have been suggested on extending configuration to resources, do not cover required Cloud properties such as elasticity or multi-tenancy.To address these limitations, we propose an approach for supporting the design and configuration of Cloud resource Allocation in configurable business process models. We target to (1) define a unified and formal description for the resource perspective, (2) ensure a correct, free-of-conflict and optimized use of Cloud resource consumption, (3) assist process providers to design their configurable resource allocation in a fine-grained way to avoid complex and large results, and (4) optimize the selection of Cloud resources with respect to the requirements related to Cloud properties (elasticity and shareability) and QoS properties.To do so, we first suggest a semantic framework for a semantically-enriched resource description in business processes aiming at formalizing the consumed Cloud resources using a shared knowledge base. Then, we build upon social business processes to provide strategies in order to ensure a controlled resource allocation without conflicts in terms of resources. Next, we propose a novel approach that extends configurable process models to permit a configurable Cloud resource allocation. Our purpose is to shift the Cloud resource allocation from the tenant side to the Cloud process provider side for a centralized resource management. Afterwards, we propose genetic-based approaches that aim at selecting optimal resource configuration in an energy efficient manner and to improve non-functional properties.In order to show the effectiveness of our proposals, we concretely developed (i) a set of proof of concepts, as a validation part, to assist the design of process models and populate a knowledge base of heterogeneous process models with Cloud resources, and (ii) performed experiments on real process models from large datasets