Online Hate Speech Research Articles

StereoHate: Toward identifying stereotypical bias and target group in hate speech detection

Abstract Though social media helps spread knowledge more effectively, it also stimulates the propagation of online abuse and harassment, including hate speech. It is crucial to prevent hate speech since it may have serious adverse effects on both society and individuals. Therefore, it is not only important for models to detect these speeches but to also output explanations of why a given text is toxic. While plenty of research is going on to detect online hate speech in English, there is very little research on low-resource languages like Hindi and the explainability aspect of hate speech. Recent laws like the “right to explanations” of the General Data Protection Regulation have spurred research in developing interpretable models rather than only focusing on performance. Motivated by this, we create the first interpretable benchmark hate speech corpus hate speech explanation (HHES) in the Hindi language, where each hate post has its stereotypical bias and target group category. Providing descriptions of internal stereotypical bias as an explanation of hate posts makes a hate speech detection model more trustworthy. Current work proposes a commonsense-aware unified generative framework, CGenEx, by reframing the multitask problem as a text-to-text generation task. The novelty of this framework is it can solve two different categories of tasks (generation and classification) simultaneously. We establish the efficacy of our proposed model (CGenEx-fuse) on various evaluation metrics over other baselines when applied to the Hindi HHES dataset. Disclaimer The article contains profanity, an inevitable situation for the nature of the work involved. These in no way reflect the opinion of authors.

“View and Hide Definitions” of Racist Hate Speech: Ethnophaulisms in Google’s English Dictionary

This paper aims to foster debate about the language of racist hate speech in online English lexicography. For this purpose, it presents a study on the treatment of ethnophaulisms, or ethnic slurs, in “powered by Oxford Languages” Google’s English dictionary. The focus is indeed on the perspective of the general user of the Internet, in light of the connection between two facets of this digital age. The first one is the strong and growing tendency among Internet users to ‘google’ their language issues. The second one is the alarming increase in cases of hate speech online, most of which are based on ethnicity and nationality, according to reports by the United Nations. Consequently, the free and pervasive content of Google’s English dictionary represents a case in point to investigate whether and how online users are warned against the power of these hate words. A selected sample of 285 English ethnic slurs have been looked up in the dictionary and, if recorded, their entries have been scrutinised to identify lexicographic data regarding their semantic relevance and offensiveness. Findings show that the majority are included, they mostly present ethnicity-related senses, but less than half of the total are treated as ethnophaulisms. In this respect, the major dictionary markers indicating offensiveness are effect labels, predominantly alone or combined with definitions. Relative to their size, thus, ethnophaulisms in Google’s English dictionary are clearly described as offensive or derogatory expressions, thus making online users aware of their hurtful nature.

Bridging the gap in online hate speech detection: A comparative analysis of BERT and traditional models for homophobic content identification on X/Twitter

Bridging the gap in online hate speech detection: A comparative analysis of BERT and traditional models for homophobic content identification on X/Twitter

Pencegahan Kejahatan Digital Dimasa Sekarang Melalui Pendidikan Kewarganegaraan

This research is aimed to find out about digital crime prevention in today's trough civic education. The method used in this research is a qualitative description of viewing data in moderation. The digital age presents a variety of ease and opportunities for society. But underneath it all, there is a dark side to be aware of, is a digital crime. Digital crime is against the laws of digital technology, such as online fraud, data theft, and hate speech. Citizenship education helps students understand the risks and dangers associated with digital crime, and it gives them the skills to protect themselves and actively participate in the digital world.Through citizen education, individuals gain a deeper understanding of their rights, obligations, honesty and responsibilities as responsible citizens when using digital technology and online ethics. With effective education, we can reduce the risk of digital crime and build a society to deal wisely with it.

Navigating the Virtual Realm of Hate: Analysis of Policies Combating Online Hate Speech in the Italian-European Context

The aim of this article is to conduct an in-depth analysis of the effectiveness of policies to counter online hate speech on social platforms. This analysis is approached through an interdisciplinary approach that combines sociological and legal perspectives to provide a non-partisan understanding of the phenomenon. The research consists of two phases: the first phase provides a general framework for the subject under examination, attempting to clearly delineate – on a definitional level – hate speech, emphasising its growing relevance in contemporary social dynamics; the second phase, focuses on an examination of the European value and regulatory framework, with particular reference to the civil law remedies that can be used in the Italian legal system to counteract the spread of online hate speech. This methodology makes it possible to identify the countervailing tools to which private individuals can resort in order to meet the requirements of justice and fairness. In this sense, the article aims to offer a broad overview of the subject matter, providing a contribution to the academic debate and to the understanding of the challenges related to the fight against online hate speech on social networks.

How harassment and hate speech policies have changed over time: Comparing Facebook, Twitter and Reddit (2005–2020)

AbstractSocial media platforms make choices about what content is and is not permissible on their platforms. For example, choices about if and how to deal with online harassment and hate speech are growing problems in many online settings. But these choices are often opaque, can vary from platform to platform, and can change over time with little notice. This study examines the ways Facebook, Twitter, and Reddit have defined harassment and hate speech, as well as who they frame as responsible for dealing with harassment and hate speech over time. Using content analysis, the policy structures that house relevant policies, the policy documents themselves, and blog posts are examined. The results illustrate a phased approach to defining harassment and hate, which has become increasingly complex and nuanced over time. Additionally, this work shows a compounding view of who is responsible, which began with users but over time has come to include the platform itself, technology, and external actors such as civil society groups. This paper highlights continued opacity and increasing complexity while also providing contextual historical information necessary for both future research and platform governance decisions.

The Virality of Hate Speech on Social Media

Online hate speech is responsible for violent attacks such as, e.g., the Pittsburgh synagogue shooting in 2018, thereby posing a significant threat to vulnerable groups and society in general. However, little is known about what makes hate speech on social media go viral. In this paper, we collect N = 25,219 cascades with 65,946 retweets from X (formerly known as Twitter) and classify them as hateful vs. normal. Using a generalized linear regression, we then estimate differences in the spread of hateful vs. normal content based on author and content variables. We thereby identify important determinants that explain differences in the spreading of hateful vs. normal content. For example, hateful content authored by verified users is disproportionally more likely to go viral than hateful content from non-verified ones: hateful content from a verified user (as opposed to normal content) has a 3.5 times larger cascade size, a 3.2 times longer cascade lifetime, and a 1.2 times larger structural virality. Altogether, we offer novel insights into the virality of hate speech on social media.

From hate speech to HateLess. The effectiveness of a prevention program on adolescents’ online hate speech involvement

Hate speech, a societal scourge, has found a potent conduit in modern technology, exacerbating its reach and impact. Adolescents are particularly vulnerable to the pernicious effects of online hate speech, which threaten their well-being and social cohesion. Addressing this critical issue, the current study uses a quasi-experimental design to evaluate the efficacy of “HateLess. Together Against Hatred” in reducing adolescents' online hate speech perpetration and victimization and increasing their online hate speech countering. A sample of 820 adolescents aged 12 to 16 (M = 13.27, SD = 1.04) from 11 German schools was divided into an intervention group (n = 567), who underwent the one-week anti-hate speech program, and a control group (n = 253). Utilizing multilevel mediation analysis, the study revealed that program participation significantly reduced online hate speech perpetration and victimization while enhancing countering hate speech. Notably, the intervention's success was partially mediated by increased empathy, which, in turn, predicted a decrease in hate speech perpetration and an increase in counter-speech, and by increased self-efficacy, which, in turn, was linked to reduced victimization and heightened countering hate speech. These promising outcomes suggest that HateLess could be a viable strategy for mitigating hate speech engagement and fostering a more respectful online discourse among adolescents.

Hate Speech Detection

Abstract: The rise in popularity of microblogging sites such as Facebook, Instagram, and Twitter has resulted in more people from different backgrounds indirectly communicating with one another. Our study aims to design an autonomous Deep Neural Network (DNN) algorithm for social media hate speech detection to tackle this problem. Using cutting-edge Natural Language Processing (NLP) techniques, the objective is to build a strong system that can recognize and categorize hate speech material in text data with accuracy. Our DNN algorithm allows for the real-time detection and moderation of offensive information, providing a proactive strategy against online hate speech. With the deployment of this technology, everyone will be able to access a safer and more welcoming online environment.

Research on Online Hate Speech Detection from Popper and Kuhn's Philosophical Perspective

Research on Online Hate Speech Detection from Popper and Kuhn's Philosophical Perspective

Who cares? How personal political characteristics are related to online counteractions against hate speech

Abstract Although many Internet users have encountered hate speech online, interventions against those incidents are unlikely. To learn more about online counteractions against hate speech, which we conceptualize as a form of online civic participation, we conducted an online survey representing adult German Internet users (N = 2,691). We investigated what distinguishes users who intervene against hate speech, focusing on associations between citizens’ prior counteractions and their personal political characteristics, their attributed responsibility for intervening against hate speech to social and political actors, and their institutional trust. Results indicate that especially Internet users’ personal political characteristics (e.g., internal political efficacy, political opinion leadership) as well as trust in regulative political institutions are positively associated with counteracting against online hate speech. While social media use and previous experience with counteractions increased the likelihood of users counteracting against hate speech, authoritarian attitudes and attributing responsibility for intervening to platform operators decreased it.

Fine-tuning Pre-trained Language Models to Detect In-game Trash Talks

Common problems in playing online mobile and computer games were related to toxic behavior and abusive communication among players. Based on different reports and studies, the study also discusses the impact of online hate speech and toxicity on players' in-game performance and overall well-being. This study investigates the capability of pre-trained language models to classify or detect trash talk or toxic in-game messages. The study employs and evaluates the performance of pre-trained BERT and GPT language models in detecting toxicity within in-game chats. Using publicly available APIs, in-game chat data from DOTA 2 game matches were collected, processed, reviewed, and labeled as non-toxic, mild (toxicity), and toxic. The study was able to collect around two thousand in-game chats to train and test BERT (Base-uncased), BERT (Large-uncased), and GPT-3 models. Based on the three models’ state-of-the-art performance, this study concludes pre-trained language models’ promising potential for addressing online hate speech and in-game insulting trash talk.

Country branding in post-truth Era: A configural narrative

This study investigates the impact of online disinformation and hate speech on a country's image, considering its implications for the destination's image and tourism prospects. Utilizing the theoretical lens of soft power, the study performs a macro-level empirical analysis across 56 countries on the influence of a diverse array of online disinformation variants and hate speech on the nation's brand image. The results from the configurational analysis demonstrate the increasing dominance of state and partisan factions-led disinformation, which collectively impact the nation's image. The findings also shed light on the diminishing role of foreign disinformation, while the ancillary function of online hate emphasizes the significance of disinformation amplified hate speech, which has become increasingly commonplace in global political discourse, with such disinformation campaigns using hate speech as an amplification strategy. The study contributes to contemporary tourism research, which underscores the vitality of the country's image to tourism and offers recommendations to destination marketing organizations towards intensifying efforts to preserve the country's image.

Hate speech detection: A comprehensive review of recent works

AbstractThere has been surge in the usage of Internet as well as social media platforms which has led to rise in online hate speech targeted on individual or group. In the recent years, hate speech has resulted in one of the challenging problems that can unfurl at a fast pace on digital platforms leading to various issues such as prejudice, violence and even genocide. Considering the acceptance of Artificial Intelligence (AI) and Natural Language Processing (NLP) techniques in varied application domains, it would be intriguing to consider these techniques for automated hate speech detection. In literature, there have been efforts to recognize and categorize hate speech using varied Machine Learning (ML) and Deep Learning (DL) techniques. Hence, considering the need and provocations for hate speech detection we aim to present a comprehensive review that discusses fundamental taxonomy as well as recent advances in the field of online hate speech identification. There is a significant amount of literature related to the initial phases of hate speech detection. The background section provides a detailed explanation of the previous research. The subsequent section that follows is dedicated to examining the recent literature published from the year 2020 onwards. The paper presents some of the hate speech datasets considered for hate speech detection. Furthermore, the paper discusses different data modalities, namely, textual hate speech detection, multi‐modal hate speech detection and multilingual hate speech detection. Apart from systematic review on hate speech detection, the paper also implement several multi‐label models to compare the performance of hate speech detection by employing classic ML technique namely, Logistic Regression and DL technique namely, Long Short‐Term Memory (LSTM) and a multiclass multi‐label architecture. In the implemented architecture, we have derived two new elements to quantify the hatefulness and intensity of hatred to improve the results for hate speech detection using Indonesian tweet dataset. Empirical Analysis of the model reveals that the implemented approach outperforms and is able to achieve improved results for the underlying dataset.

Hate speech detection in the Arabic language: corpus design, construction, and evaluation.

Hate Speech Detection in Arabic presents a multifaceted challenge due to the broad and diverse linguistic terrain. With its multiple dialects and rich cultural subtleties, Arabic requires particular measures to address hate speech online successfully. To address this issue, academics and developers have used natural language processing (NLP) methods and machine learning algorithms adapted to the complexities of Arabic text. However, many proposed methods were hampered by a lack of a comprehensive dataset/corpus of Arabic hate speech. In this research, we propose a novel multi-class public Arabic dataset comprised of 403,688 annotated tweets categorized as extremely positive, positive, neutral, or negative based on the presence of hate speech. Using our developed dataset, we additionally characterize the performance of multiple machine learning models for Hate speech identification in Arabic Jordanian dialect tweets. Specifically, the Word2Vec, TF-IDF, and AraBert text representation models have been applied to produce word vectors. With the help of these models, we can provide classification models with vectors representing text. After that, seven machine learning classifiers have been evaluated: Support Vector Machine (SVM), Logistic Regression (LR), Naive Bays (NB), Random Forest (RF), AdaBoost (Ada), XGBoost (XGB), and CatBoost (CatB). In light of this, the experimental evaluation revealed that, in this challenging and unstructured setting, our gathered and annotated datasets were rather efficient and generated encouraging assessment outcomes. This will enable academics to delve further into this crucial field of study.

Online hate speech victimization: consequences for victims’ feelings of insecurity

This paper addresses the question whether and to what extent the experience of online hate speech affects victims’ sense of security. Studies on hate crime in general show that such crimes are associated with a significantly higher feeling of insecurity, but there is little evidence concerning feeling of insecurity due to online hate speech. Based on a secondary data analysis of a representative population survey in Lower Saxony, Germany, on the topic of cybercrime in 2020 (N = 4,102), we tested three hypotheses regarding the effect of offline and online hate speech on feelings of insecurity. As a result, compared to non-victims, victims of online hate speech exhibit a more pronounced feeling of insecurity outside the Internet, while victims of other forms of cybercrime do not differ in this regard from non-victims. We found no effect for offline hate speech when relevant control variables were included in the statistical model. Possible reasons for this finding are assumed to lie in the characteristics of the phenomenon of online hate speech, for example, because the hateful content spreads uncontrollably on the Internet and reaches its victims even in protected private spheres.

The ecology of online hate speech: Mapping expert perspectives on the drivers for online hate perpetration with the Delphi method

AbstractOnline hate speech (OHS) is a prevalent issue on social media with harmful consequences for individuals and society. Counteracting this issue requires a better understanding of the factors that drive behaviors such as posting hate speech or making hateful comments. The present study contributes to this need by investigating the personal, social, and digital determinants for OHS perpetration. By conducting a two‐round Delphi study, data was collected from an interdisciplinary sample of experts. Practitioners were included in this sample to bridge gaps between research and fieldwork on OHS. The first survey collected information through open‐ended questions, which was then organized in different themes and lists of determinants in a thematic analysis. These determinants were rated for relevance in a follow‐up survey. Data analysis of the second survey used nonparametric statistics. From a total list of 129 determinants, the participants reached a consensus on 27 determinants as most relevant (median > 4) for facilitating the production and dissemination of OHS. An ecological perspective was applied to interpret these findings. As a result, we propose an ecological model of personal, social, and digital determinants for online hate perpetration to illustrate their influence on individual behavior as well as their possible connections. The model provides a comprehensive overview of factors facilitating OHS perpetration, which can help to identify points for intervention.

Comparative Analysis of NLP Techniques for Hate Speech Classification in Online Communications

This research aimed to compare the effectiveness of two Natural Language Processing (NLP) techniques—SpaCy's word embeddings and Sklearn's TF-IDF vectorization—in identifying hate speech within online comments. Utilizing a balanced dataset, each model was meticulously assessed on its ability to classify comments as 'hateful' or 'non-hateful'. The evaluation metrics employed were precision, recall, F1-score, and overall accuracy. The model using SpaCy's word embeddings achieved an accuracy of 65%, with equal precision and recall for both classes. The Sklearn's TF-IDF vectorization model, however, demonstrated superior performance with an overall accuracy of 75% and an enhanced ability to correctly identify hateful comments, evidenced by a 77% recall rate. This suggests that the TF-IDF model is more adept at discerning nuanced expressions of hate speech. The study's findings highlight the critical role of vectorization methods in the field of automated content moderation and stress the importance of continued innovation and model adaptation to effectively manage the evolving nature of online hate speech.

Hate Speech Detection in Limited Data Contexts Using Synthetic Data Generation

A growing body of work has focused on text classification methods for detecting the increasing amount of hate speech posted online. This progress has been limited to only a select number of highly resourced languages causing detection systems to either under-perform or not exist in limited data contexts. This is mostly caused by a lack of training data, which are expensive to collect and curate in these settings. In this work, we propose a data augmentation approach that addresses the problem of lack of data for online hate speech detection in limited data contexts using synthetic data generation techniques. Given a handful of hate speech examples in a high-resource language such as English, we present three methods to synthesize new examples of hate speech data in a target language that retains the hate sentiment in the original examples but transfers the hate targets. We apply our approach to generate training data for hate speech classification tasks in Hindi and Vietnamese. Our findings show that a model trained on synthetic data performs comparably to, and in some cases outperforms, a model trained only on the samples available in the target domain. This method can be adopted to bootstrap hate speech detection models from scratch in limited data contexts. As the growth of social media within these contexts continues to outstrip response efforts, this work furthers our capacities for detection, understanding, and response to hate speech. Disclaimer: This work contains terms that are offensive and hateful. These, however, cannot be avoided due to the nature of the work.

‘You Got My Back?’ Severity and Counter-Speech in Online Hate Speech Toward Minority Groups

ABSTRACT Minority members are often targeted by online hate speech. In this article, we advance the literature by conceptualizing online hate speech as a social identity threat, whose consequences can adversely affect intergroup relations at large. Within two experiments, we examine the effects of homophobic (Study 1, N = 225) and racist (Study 2, N = 194) hate speech on target minority members’ attitudes toward the majority group and intentions to avoid it. In addition, we test the potential mitigating role of counter-speech presence and source (minority vs. majority member). Drawing from theoretical accounts of social identity and perceived opinion climate, we also shed light on novel underlying mechanisms that explain the effect of online hate speech on intergroup relations: perceived identity threat (PIT) and perceived negative representation (PNR). Across both studies, more severe hate speech resulted in greater PIT and PNR, which was related to more hostile attitudes toward the majority group and/or intentions to avoid majority members. Counter-speech was largely ineffective in alleviating these negative outcomes. However, Study 2 revealed that counter-speech from a majority member could mitigate PIT, but only for less severe forms of hate speech. We discuss theoretical and practical implications for research in media psychology.