Skip to main content
main-content
Top

About this book

This book provides an overview of fake news detection, both through a variety of tutorial-style survey articles that capture advancements in the field from various facets and in a somewhat unique direction through expert perspectives from various disciplines. The approach is based on the idea that advancing the frontier on data science approaches for fake news is an interdisciplinary effort, and that perspectives from domain experts are crucial to shape the next generation of methods and tools.
The fake news challenge cuts across a number of data science subfields such as graph analytics, mining of spatio-temporal data, information retrieval, natural language processing, computer vision and image processing, to name a few. This book will present a number of tutorial-style surveys that summarize a range of recent work in the field. In a unique feature, this book includes perspective notes from experts in disciplines such as linguistics, anthropology, medicine and politics that will help to shape the next generation of data science research in fake news.
The main target groups of this book are academic and industrial researchers working in the area of data science, and with interests in devising and applying data science technologies for fake news detection. For young researchers such as PhD students, a review of data science work on fake news is provided, equipping them with enough know-how to start engaging in research within the area. For experienced researchers, the detailed descriptions of approaches will enable them to take seasoned choices in identifying promising directions for future research.

Table of Contents

Frontmatter

A Multifaceted Approach to Fake News

Abstract
This introductory chapter provides the contextual basis and inspiration for this volume/book, as well a summary of the chapter contents. We believe that the multifaceted approach to fake news we have used in this book is quite unique and hopefully will open up fresh perspectives and questions for the interested reader. In this chapter, we start by motivating the need for such a multifaceted approach toward fake news, followed by introducing the two-part structure of this book: surveys and perspectives. Further, we briefly describe the various surveys and perspectives that we cover in the subsequent chapters in this book.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Survey

Frontmatter

On Unsupervised Methods for Fake News Detection

Abstract
In this chapter, we consider a reasonably underexplored area in fake news analytics, that of unsupervised learning. We intend to keep the narrative accessible to a broader audience than machine learning specialists and accordingly start with outlining the structure of different learning paradigms vis-à-vis supervision. This is followed by an analysis of the challenges that are particularly pertinent for unsupervised fake news detection. Third, we provide an overview of unsupervised learning methods with a focus on their conceptual foundations. We analyze the conceptual bases with a critical eye and outline other kinds of conceptual building blocks that could be used in devising unsupervised fake news detection methods. Fourth, we survey the limited work in unsupervised fake news detection in detail with a methodological focus, outlining their relative strengths and weaknesses. Lastly, we discuss various possible directions in unsupervised fake news detection and consider the challenges and opportunities in the space.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Multi-modal Fake News Detection

Abstract
The primary motivation behind the spread of fake news is to convince the readers to believe false information related to certain events or entities. Human cognition tends to consume news more when it is visually depicted through multimedia content than just plain text. Fake news spreaders leverage this cognitive state to prepare false information in such a way that it looks attractive in the first place. Therefore, multi-modal representation of fake news has become highly popular. This chapter presents a thorough survey of the recent approaches to detect multi-modal fake news spreading on various social media platforms. To this end, we present a list of challenges and opportunities in detecting multi-modal fake news. We further provide a set of publicly available datasets, which is often used to design multi-modal fake news detection models. We then describe the proposed methods by categorizing them through a taxonomy.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Deep Learning for Fake News Detection

Abstract
The widespread usage of fake news through social media outlets causes unpleasant societal outcomes. The research efforts to automatically detect and mitigate its use are essential because of their potential to influence the information ecosystem. A vast amount of work using deep learning techniques paved a way to understand the anatomy of fake news and its spread through social media. This chapter attempts to take stock of such efforts and look beyond the possibilities in this regard. The focus is given mainly to deep learning models and its use in fake news detection. A comprehensive survey of the current literature and datasets used, along with evaluation metrics, are highlighted. Finally, promising research directions toward fake news detection are mentioned.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Dynamics of Fake News Diffusion

Abstract
Modeling information diffusion on social media has gained tremendous research attention in the last decade due to its impact in understanding the overall spread of news contents through network links such as followers, friends, etc. Those fake stories which gain quick visibility are deployed on social media in a strategic way in order to create maximum impact. In this context, the selection of initiators, the time of deployment, the estimation of the reach of the news, etc. play a decisive role to model the spread appropriately. In this chapter, we start by defining the problem of fake news diffusion and addressing the challenges involved. We then model information cascade in various ways such as a diffusion tree. We then present a series of traditional and recent approaches which attempt to model the spread of fake news on social media.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Neural Language Models for (Fake?) News Generation

Abstract
Recent progress in natural language processing (NLP), in conjunction with deep neural network models, has created a broad interest in natural language generation and many downstream applications of NLP. Deep learning models with multitudes of parameters have achieved remarkable progress in machine-generated news items indistinguishable from human experts’ articles. Though the developed techniques are for authentic text generation and entertainment purposes, its potential use in social media for propaganda, defamation, and misinformation propagation poses a considerable challenge to the research community. This chapter attempts to present a study on various pre-trained neural models for natural language processing in general and their potential use in news generation. While showing these models’ limitations, the chapter describes the future works in the NLP domain on language generation.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Fact Checking on Knowledge Graphs

Abstract
Fact checking, which verifies whether a given statement is true, could play a vital role in fake news detection. For example, for a given piece of news, a potential solution could involve a series of steps, including extracting statements from the news via text parsing, checking the validity of the extracted statements (i.e., fact checking), and classifying the news as fake if some statements have been confirmed to be false and performing further fake news detection processes otherwise. Considering that knowledge graphs are a popular way of representing knowledge, which could be used for verifying or counter-verifying statements, several solutions have been proposed that make use of knowledge graphs for fact checking. In this chapter, recent studies on fact checking with the help of knowledge graphs are reviewed, and three representative solutions, namely, Knowledge Linker, PredPath, and Knowledge Stream, are introduced with some details. Specifically, Knowledge Linker utilizes the semantic proximity metrics for mining knowledge graphs, PredPath employs the link prediction method and introduces a newly defined metric, and Knowledge Stream models the fact-checking problem as an optimization problem and uses flow theory for solving the problem.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Graph Mining Meets Fake News Detection

Abstract
Nowadays, the diversified services on social media make news diffused at higher rate and larger volumes, which poses unique challenges in terms of the efficiency, scalability, and accuracy on the fake news detection. To solve these issues, graph mining, as a promising direction of data mining, has successfully attracted attentions of recent studies. In this chapter, we present a comprehensive study on recent graph-based fake news detection approaches and show how graph mining enables the whole task. We first introduce different kinds of information related to fake news, then divide the existing graph-based approaches into two scenarios, where various graphs and graph patterns are introduced to model the information on social media and characterize features of the fake news, respectively.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Perspectives

Frontmatter

Fake News in Health and Medicine

Abstract
With the rise of social media, the world is faced with the challenge of increasing health-related fake news more than ever before. We are constantly flooded with health-related information through various online platforms, many of which turn out to be inaccurate and misleading. This chapter provides an overview of various health fake news and related studies which have been reported in various news articles and scientific journals. Some of the studies conducted on health misinformation identified a prominence of vaccine- and cancer-related fake news. The popularity of so-called unproven natural cures for cancer and other diseases is alarming. The chapter also highlights the importance of maintaining accurate and effective scientific communication in this COVID-19 pandemic-hit world to safeguard public health. The current pandemic has also proved fertile ground for spreading misinformation. The chapter brings the audience’s attention to the consequences of health misinformation, ranging from giving false hope to patients to the hurdles it poses to effective medical care. Finally, the chapter addresses some of the possible strategies to keep health misinformation in check.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Ethical Considerations in Data-Driven Fake News Detection

Abstract
Data-driven and AI-based detection of fake news has seen much recent interest. The focus of research on data-driven fake news detection has been on developing novel and effective machine learning pipelines. The field has flourished with the rapid advances in deep learning methodologies and the availability of several labelled datasets to benchmark methods. While treating fake news detection as yet another data analytics problem, there has been little work on analyzing the ethical and normative considerations within such a task. This work, in a first-of-its-kind effort, analyzes ethical and normative considerations in using data-driven automation for fake news detection. We first consider the ethical dimensions of importance within the task context, followed by a detailed discussion on adhering to fairness and democratic values while combating fake news through data-driven AI-based automation. Throughout this chapter, we place emphasis on acknowledging the nuances of the digital media domain and also attempt to outline technologically grounded recommendations on how fake news detection algorithms could evolve while preserving and deepening democratic values within society.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

A Political Science Perspective on Fake News

Abstract
Contemporary concerns about “fake news” are typically framed around the need for factual accuracy, accountability, and transparency in public life at both national and international level. These are long-standing concerns within political science, but the problem of “fake news” and its associated impact on the fundamental political questions about who governs and how have taken on new potency in the digital age. In this chapter, we begin by considering what is meant by fake news before examining the issue in a historical political context. The chapter then turns to more recent manifestations of fake news and the real-world challenges it presents. A final section considers how fake news has attracted interest in the study of elections and voting behaviour, international relations and strategic narratives, and transparency and trust in government.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Fake News and Social Processes: A Short Review

Abstract
The explosive growth in social media, social networking, and messaging platforms has seen the emergence of many undesirable social phenomena. A common thread among many of these social behaviors is disinformation propagation, through falsehoods of many shades and grades that are quickly propagated to millions of people. In this chapter, we focus on disinformation propagation mainly in the garb of fake news, which contains deceptive, distorted, malicious, biased, polarizing, inaccurate, unreliable, unsubstantiated, and unverified or completely false or fabricated information. We examine the literature related to the sociological analysis of the fake news phenomenon and its impact on social processes such as elections and vaccination. We also outline directions for further research.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Misinformation and the Indian Election: Case Study

Abstract
Diverse demographics, culture, and language; a troubled history of communal violence, polarised politics, and sensationalist media; and a recent explosion in smartphone ownership and Internet access have created a “fake news” crisis in India which threatens both its democratic values and the security of its citizens. One of the unique features of India’s digital landscape is the prevalence of closed networks – ideologically homogeneous groups of individuals communicating on private platforms – in which misinformation proliferates. This poses several challenges: the encryption of private messages makes tracking and analysing the spread of information through these channels difficult; accusations of censorship and surveillance can prevent governments from tackling misinformation propagated through private groups; ethical considerations associated with the extraction of data from encrypted, private conversations; highly influential means of disseminating information, with users receptive to messages which fit the common worldview of the group; and speed of proliferation, i.e. information spread to a large user base at the touch of a button. In this chapter, we showcase some of our core technologies in the areas of credibility and veracity assessment. Our key findings during the Indian general election 2019 indicate that individual users play a major role in the solution to fake news and misinformation. Adopting content verification strategies even by 1% of WhatsApp users will help vaccinate India’s WhatsApp networks against fake news. Whilst volume must increase, the speed of fact-checks is the vital improvement which the fact-checking industry must undergo – semi-automated fact-checks proved 35 times more effective than traditional fact-checks in fighting fake news on WhatsApp – to debunk rumours early in their propagation path and spread fact-checks to audiences that came across the original piece of problematic content.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

STS, Data Science, and Fake News: Questions and Challenges

Abstract
Fake news has become the norm in our times. With the coming of social media platforms, where anyone can write/post on any issues without any regulations, sometimes based on what they read from various platforms and sometimes based on what they are asked to believe in by political ideologies, fake news has become a phenomenon that requires serious academic investigation as it has dangerous consequences in society. This chapter attempts to argue for a possible and productive conversation between science and technology studies (STS) and data science to talk about the politics of fake news production. It argues that STS can work as a close ally of data science to bring in questions of power and politics associated with fake news, and its methods can be used in data science to make it more socially relevant.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G

Linguistic Approaches to Fake News Detection

Abstract
To date, there is no comprehensive linguistic description of fake news. This chapter surveys a range of fake news detection research, focusing specifically on that which adopts a linguistic approach as a whole or as part of an integrated approach. Areas where linguistics can support fake news characterisation and detection are identified, namely, in the adoption of more systematic data selection procedures as found in corpus linguistics, in the recognition of fake news as a probabilistic outcome in classification techniques, and in the proposal for integrating linguistics in hybrid approaches to fake news detection. Drawing on the research of linguist Douglas Biber, it is suggested that fake news detection might operate along dimensions of extracted linguistic features.
Deepak P, Tanmoy Chakraborty, Cheng Long, Santhosh Kumar G
Additional information