Fake news is now viewed as one of the greatest threats to democracy and journalism [17]. The reach of fake news was best highlighted during the critical months of the 2016 U.S. presidential election campaign, where top twenty frequently-discussed false election stories generated 8,711,000 shares, reactions, and comments on Facebook, ironically, larger than the total of 7,367,000 for the top twenty most-discussed election stories posted by top major news websites [11]. Our economies are not immune to fake news either, impacting stock markets and leading to massive trades. For example, fake news claiming that Barack Obama was injured in an explosion wiped out $130 billion in stock value [5, 14].
The generous benefits in fake news activities are one of the motivations for people to initiate and engage in such activities. Consider dozens of “well-known” teenagers in the Macedonian town of Veles who posted fake news for millions on social media and became wealthy by penny-per-click advertising during the 2016 U.S. presidential election [12]. Such stories attach greater importance to fake news detection and intervention as they provide an incentive for individuals to become the next “Macedonian teenagers” in the upcoming elections all around the world. With fake news detection research in its early stages, greater opportunities exist for such malicious individuals to create and spread fake news in the absence of a worry. On the other hand, it has been suggested that fake news is difficult to be recognized by the pubic, which leads to unintentional engagement in spreading fake news [17]; studies in social psychology and communications have demonstrated that human ability to detect deception is slightly better than chance, with a mean accuracy rate of 54 % in over 100 experiments [8]. Such difficulty is also related to how individuals adjust (or correct) their judgments to fake news when it has already gained their trusts [7].
Facing such grim situation, this tutorial aims to (i) provide a clear understanding of fake news; (ii) attract researchers within general areas of data/graph mining, machine learning, Natural Language Processing (NLP), and Information Retrieval (IR) to conduct research on fake news and its detection and further facilitate its development; and (iii) encourage a collaborative effort of experts in computer and information science, political science, journalism, social science, psychology and economics to work on fake news detection, where such efforts can lead to fake news detection that is not only highly efficient, but more importantly, interpretable [9]. The tutorial contains the following four parts to achieve these goals:
I. Fake News and Related Concepts. We first present two definition of fake news in a broad and narrow way, which enables one to define fake news in terms of three general characteristics: (i) information authenticity, (ii) author intention, and (iii) whether the given information is in form of news. Such characteristics help differentiate fake news from the truth, as well as from several common related concepts, e.g., mis-/dis-information, satire news, and rumors. We will specify why fake news is defined in such ways, what each characteristic indicates, and how it can be evaluated, quantified, or used to differentiate fake news from related concepts.
II. Fundamental Theories. Human vulnerability to fake news, which can bring in useful clues or further complicate fake news detection, has been a subject of interdisciplinary research [18]. For instance, achievements in forensic psychology such as Undeutsch hypothesis [13] have pointed out the style and quality differences between the truth and deceptive information. Similarly, interdisciplinary research has looked at why individuals spread fake information, considering that the borderline between malicious and normal users becomes unclear – normal people can also frequently and unintentionally participate in fake news activities, e.g., due to their social identity [1] or preexisting knowledge [4]. This tutorial conducts a comprehensive cross-disciplinary survey of literature on such theories. We review more than twenty well-known theories that can contribute to our understanding of fake news and participants in fake news activities [17]. We present and discuss the problems arising as explained by these theories, ranging from the patterns they can reveal, the qualitative and quantitative fake news studies one can conduct based on these studies, to the specific roles they can play in detecting fake news.
III. Detection Strategies. Detecting fake news is a complex and multidimensional task: it involves assessing multiple characteristics of news such as its authenticity, author intention, and its literary form. Furthermore, fake news is formed by multiple components (e.g., headline, body text, attached image(s)), and available information on fake news that can be utilized in predicting fake news sharply increases as it starts to disseminate online (e.g., feedback from users such as comments, its propagation paths on networks and its spreaders). Such components and information can be in the form of text, multimedia, network, etc., corresponding to various applicable techniques and usable resources.
To methodically and comprehensively present the ways to detect fake news, in this tutorial, we will specify how fake news detection can be conducted respectively from four perspectives (i.e., knowledge, style, propagation and credibility) – their corresponding general strategies, targeting fake news characteristic that can be evaluated, components and information that can be utilized, applicable techniques, and some typical approaches.
Generally speaking, fake news detection from a knowledge perspective is a “comparison” between the relational textual knowledge extracted from to-be-verified news articles and that of knowledge graphs representing facts or ground truth [2, 6]. The construction of knowledge graphs is an active research area within IR. Such “comparison” is often reduced to a link prediction (or knowledge inference) task, which directly evaluates news authenticity. Style- based fake news detection aims to capture the differences in writing styles between fake and true news, which often relies on NLP techniques and is conducted within a machine learning framework. News style can be extracted from the text [16], images [15], and/or videos within to-be-verified content, enabling one to indirectly evaluate the intention of the creator of news articles. Propagation- based and credibility-based fake news detection both further exploit information provided in news propagation on social media, where the former mainly relies on news cascades or self-defined graphs [14], while the latter emphasizes on exploring the credibility relationships between news articles and entities such as clickbait, publishers, spreaders, comments, etc. [3, 10]. Hence, research tasks involved can be correlated to clickbait detection, opinion spam detection, and the like. Here, graph optimization algorithms often play an important role to solve the target problems.
IV. Open Issues. In the final section of the tutorial, we will present the challenges and open issues that are important but have not been addressed (or thoroughly addressed) in current studies. Such challenges and open issues are three-fold: (i) challenges brought from news characteristics, e.g., the timeliness of news articles demands real-time knowledge graphs that can assure knowledge timeliness; (ii) open issues attached to model explainability; and (iii) open issues attached to model performance, e.g., the completeness of knowledge graphs and cross-domain generalization of style-based approaches. Five tasks, namely fake news early detection, checkworthy content identification, cross-domain/topic/language study of fake news, representation learning for fake news detection, and fake news intervention) will be thus highlighted, with discussions on why these tasks are crucial and potential ways to address each task.
It can be argued that it was precisely in the context of an aggravated informational confrontation involving several international actors of political activity at the state level (Russia, the USA, Ukraine, the EU countries) that “fake journalism” was almost officially legalized in media activities.
At one time, Yu.V. Klyuyev in the monograph “Political Discourse in Mass Communication: An Analysis of Public Political Interaction” [7] convincingly showed that the nature of statements by the media in the current system of media coordinates is determined by many influence factors. But first of all, the fact that the subjects of this process (which are both the media themselves and the journalists working for them) may have a certain position on a particular issue of the public political agenda. It was her who, as the experience of reporting on the events surrounding the Ukrainian crisis showed during the foreseeable period (November 2013 – summer 2015), was defended and will be protected by these entities by hook or by crook. And fake, as a specific format for working with information and its sources, in this way becomes an unexpected and effective tool for political struggle from a purely entertaining, post-modern fun or game of Internet fans.
Many experts and media researchers, as well as political scientists, have already drawn attention to the fact that in the Ukrainian information discourse, in addition to the dominant political theme, there is also a clear priority for using network sources of video information. And those, in turn, attract all available developments related to the IT world as technological assistance. It is difficult to disagree with the same D. Dragunsky, who rightly remarked: “The digital revolution made proving any fact very probable. This is due to both the features of digital editing and the practical immensity of resources. You can give one hundred counterproflinks to each prooflink, and so on. This opens up hitherto unprecedented opportunities for malicious fraud, and for postmodern games, and the difference between the first and second is not always obvious. And further – since the difference between a disinterested game and intentional falsification is not clear, the difference between a fake and a fact as such is gradually erased “[5, p. nine].
The world audience could see a lot of examples of such “fakes” that were used to cover certain political aspects of Ukrainian events since the beginning of their time: from the “picture” of supposedly Russian tanks in Ukraine, borrowed from the popular a computer game, before the mythological fake about the allegedly “crucified by Ukrainian nationalists” boy in the village he had seized.Similar methods of obviously fake origin are rapidly gaining political weight, because they are used most often for provocative purposes with a clear desire to politically aggravate the situation around Ukraine at one stage or another, the development of the conflict.
However, the mass consciousness in the Western world in itself is filled with such convincing fakes that sometimes it seems there is no need to create new ones. In a sense, such a powerful “dream factory” as Hollywood is partly involved in their construction. However, even in this holy of holies of American mass culture, there are still creative individuals who reveal the mechanisms for creating such fakes of a world scale. It’s enough to recall the ostrogrotsikovskoy picture director Barry Levinson’s “The tail wags the dog (Dodger)”, released in 1997, shortly before the Clinton-Lewinsky scandal and the start of the NATO bombing of Yugoslavia. However, it was in this film that all the tricks that were later used in reality by politicians and the media that accompanied them were predicted. In the Levinson film, in order to distract the attention of the American audience from the sexual scandal in which the unnamed President of the United States was involved, specially hired specialists from the world of show business arrange ... a virtual war with Albania. The reason for the promotion of this information trend is the story taken in the pavilions of one of the television studios about allegedly documentary footage of a rescuing refugee fleeing a village allegedly captured by Albanian terrorists. However, it is worth noting one significant circumstance, even if it relates to the sphere of fiction in the film “The tail wags the dog” (the name is so symbolic!). Explicit infofake served as evidence for a political decision. Another thing is that the “war with Albania” was also a grand mystification. The fake format worked and gave concrete results, which the film was perceived by the vast majority of American society.
It is unlikely that the audience of the premiere of this tape in 1997 could have suggested that the director’s fiction would turn into a fake reality of the current media environment. It is worth considering the significant difference that distinguishes the information space of the end of the past century from the current one, functioning under conditions of dominance of the principles of show civilization. The availability of wide access to network resources in modern conditions has significantly changed the paradigm of reliability obtained from the virtual space of the Information Network. Surprisingly, but true.Most journalists almost ignore the possibility of falsification or mystification of the information received and “visual” video evidence, thereby opening the gates of the media space for the penetration of fakes of various kinds and meanings.
An example of such a situation is associated with the possible re-enactment of those video-punishments organized by ISIS militants over captured prisoners and hostages. As you know, captured on video footage of brutal massacres distributed exclusively through the Internet. And only then they were reproduced on the air of the largest and most respected television companies, and also posted on the websites of news agencies, which could not but lead to a wave of indignation among the world community and to the requirements of more decisive action in relations of a self-proclaimed Islamic state.
It is significant that the fact of a possible forgery was also detected using the corresponding video in the same network. And again, we have to admit that it is similar, untested for authenticity video information that is rather an illustration in the fake format, the main structure-forming characteristic of which is the deliberate misleading of the audience of this fake. By the way, as recent international political experience has shown, those who posted such videos reached their goal: against ISIS, at least some military actions actually started. However, the question of authorship and the place of creation of such bloody “video fakes” still remains an open question.
The purpose of using this format can be any. But in the context of the current informational confrontation between various subjects of the media space, most often it turns out to be political. As for the meaning of using fakes in a particular political situation, then a certain scientific and expert opposition to them may be to find an answer to the question of who is benefiting from this. In addition, in order to prevent the “fakeization” of modern media space, it is necessary to develop theoretical and practical tools to combat this format. And this, in turn, puts forward on the scientific and methodological agenda the issue of principles for determining the reliability of information received by journalists and the media.