key: cord-0141408-5v91nevy
authors: Leitner, Stephan; Gula, Bartosz; Jannach, Dietmar; Krieg-Holz, Ulrike; Wall, Friederike
title: Infodemics: A call to action for interdisciplinary research
date: 2020-07-23
journal: nan
DOI: nan
sha: b67eaef6982a45225485a226931164873fb18b33
doc_id: 141408
cord_uid: 5v91nevy

Research on infodemics, i.e., the rapid spread of (mis)information related to a hazardous event such as the COVID-19 pandemic, requires the integration of a multiplicity of scientific disciplines. The dynamics emerging from infodemics have the potential to generate complex behavioral patterns. For the field of Business and Economics, understanding these dynamics is of ultimate importance: it supports, for example, anticipating individual behavior, which might help reduce the uncertainty entailed by the COVID-19 pandemic and allows for assessing the efficiency of policy decisions to contain its effects. In addition to the field of Business and Economics, we take into account the following disciplines: Through the lens of Computer Science and Information Systems, the information accessible to individuals is central, whereby the way information spreads in a society is strongly affected by the employed algorithms for information provision and by personalization. From the perspective of Linguistics, specific language signals in communication which emerge during pandemics have to be taken into account (e.g., emotion-related words, avoiding causal terms). Considering linguistic patterns in the context of infodemics appears to be highly relevant as they strongly affect how information is interpreted, fact-checked, made sense of by non-expert persons, and the way misinformation is automatically detected. From a Cognitive Psychology point of view, the focus is on how motives, intuition and affect influence the search and evaluation of information, and on how cognitive processes, the digital information environment and linguistic patterns together shape individuals' understanding of critical events, risk perception and behavior. The perspective of Business and Economics allows for integrating these perspectives into the wider context of economic systems (e.g., organizations or the society).

In the past few months, we have all witnessed the short-term responses to the COVID-19 pandemic, such as extensive lockdowns in both the private and the professional spheres. We have experienced a multiplicity of near-term consequences of such a hazardous event, for example the enormous costs in lives and the substantial increase in uncertainty related to the capacity of healthcare systems. Alongside the COVID-19 pandemic, we can also observe an accompanying infodemic: Information and misinformation related to the pandemic spread rapidly, which makes it difficult for decision-makers to find reliable sources of information. This can have strong and often negative effects on individual behavior and on the efficiency of counter-measures deployed by policy-makers (World Health Organization 2018). A prominent example for the rapid spread of (mis)information is the extensive promotion of hydroxychloroquine as treatment for the SARS-CoV-2 virus in social media without reliable evidence of its efficacy and secondary effects, which was followed by an emergency use authorization by the US Food and Drug Administration (FDA) and a significant increase in prescriptions (Baker et al. 2020a; Lenzer 2020) . Later, this authorization was revoked by the FDA as no benefit for decreasing the likelihood of death or recovery by treating patients with hydroxychloroquine could be observed, whereas serious side effects such as heart rhythm problems, kidney injuries, and liver problems were reported (U.S. Food & Drug 2020).

Aside from these short-term consequences, long-term effects can be expected in the future. These long-term effects might include major impacts on organizations and markets which might, for example, be caused by adaptations in consumer and travel behavior (Fernandes 2020; Leitner 2020; Nicola et al. 2020) . As the pandemic and the infodemic are still ongoing, the scale of the effects, however, is unknown (Baker et al. 2020b; Goodell 2020) . In order to react appropriately to the current situation, it is of ultimate importance for the fields of Business and Economics to understand the dynamics which emerge from infodemics. The ongoing digitalization makes infodemics a particularly rising challenge for researchers as the popular use of digital communication technologies and social media accelerates the diffusion of information substantially: This trend is also reflected in the spread of (mis)information related to the COVID-19 pandemic (Pulido et al. 2020; Rovetta and Bhagavathula 2020) ; it has led to significantly more worldwide fear, panic, and uncertainty when compared to infodemics that emerged from previous pandemics (Cinelli et al. 2020; Vaezi and Javanmard 2020) .

Dealing with infodemics requires a highly interdisciplinary approach and this position paper is a call to action for such interdisciplinary research on infodemics. In order to highlight the importance of interdisciplinary collaboration on this topic, we particularly take into account the following four perspectives: First, through the lens of Computer Science and Information Systems, the information accessible to individuals is central, whereby the way information spreads in a society is strongly affected by the employed algorithms for information provision and by the way information provision is aligned with personal interests (see Sec. 2.1). Second, from the perspective of Linguistics, specific language signals in communication which emerge during pandemics have to be taken into account (e.g., emotion-related words, avoiding causal terms). Considering linguistic patterns in the context of infodemics is highly relevant as these patterns strongly affect the way information is interpreted, fact-checked, made sense of by non-expert persons, and how misinformation is automatically detected (see Sec. 2.2). Third, from a Cognitive Psychology point of view, the focus is on the assessment of how available information (e.g., driven by algorithms and by personalization) and the language signals affect the way pandemics and policies are perceived and on how they influence individual behavior (see Sec. 2.3). Fourth, while the Cognitive Psychology viewpoint mainly focuses on the level of the individual, the perspective of Business and Economics allows for integrating the aforementioned perspectives into the wider context of economic systems (e.g., organizations or the society) (see Sec. 2.4) . By doing so, the holistic and integrated perspective prepares the grounds for gaining fundamental insights into the dynamics emerging from infodemics.

Selected perspectives on infodemics

In the last two decades, we have observed disruptive changes in terms of how information is spread in societies. While traditional mass media channels continue to exist, the Web and in particular Social Media have become the main source of information for the majority in today's digital society. These developments have made information not only more easily accessible for consumers but have also led to the rise of digital businesses like Google, Facebook, or Twitter, which have an enormous, global-scale reach. As a result, information that is spread via such channels is often viewed by millions of people within short periods of time.

Unlike traditional media, such online information channels do not rely on journalists or editors to ensure the quality or reliability of the information that is spread. This makes such channels generally vulnerable of being used for the dissemination of misinformation in the form of an infodemic. A particular problem that can additionally contribute to the rapid spread of misinformation on such channels lies in the fact that the selection of the content provided through these channels is determined by algorithms. These algorithms are usually optimized to identify, in a personalized way, those pieces of content that are most likely viewed by an individual consumer. The underlying reason for trying to optimize the number of 'clicks' instead of finding the most relevant pieces of content for a user lies in the advertisement-based business model of many online services.

The algorithms that select and rank the content for the individual users, e.g., in the form of recommender systems (Jannach et al. 2010) , are in most cases based on statistics and machine learning. As such, they learn over time which pieces of content, e.g., news articles or videos, are likely to be clicked on by users. As a result of using such learning strategies, a reinforcement ("blockbuster") effect can often be observed (Fleder and Hosanagar 2009) where the rich get richer: Content that has reached a certain level of popularity continues to be recommended to even more users.

But this reinforcement bias is not the only problem of today's algorithm-controlled dissemination of information. The increased levels of personalization can also lead to filter bubbles (Pariser 2011) and echo chambers (Celis et al. 2019 ). The reason is that modern algorithms learn over time which type of content a certain user likes or dislikes. Again, to optimize their business, the algorithms will try to focus on content that the user is likely to click on. In the context of an infodemic, this can lead to the effect that users at some stage are presented only with one of several existing theories, e.g., regarding the appropriateness of certain measures taken by policymakers to contain the pandemic. This one-sided information state, as a result, can further reinforce the spread of misinformation.

In the academic literature in the fields of Computer Science and Information Systems, these potentially negative effects of personalization and algorithm-driven content recommendations have recently attracted increased research interest. Correspondingly, questions of algorithmic "fairness" and how to provide users with understandable explanations why certain items were recommended moved into the focus of researchers. These problems are, however, far from being solved. Besides algorithmic challenges, a fundamental problem usually lies in defining what being fair actually means in a given context (Friedler et al. 2016; Burke 2017; Abdollahpouri et al. 2020 ).

While Computer Science algorithms distribute containers of information fitting systematic behavioral patterns (e.g., counting clicks as indicators of content preferences and recommending "similar" containers), the study of language (Linguistics) is directly concerned with the contents of these information containers -at the lexical level of words, at the compositional level of utterances (assertions, claims, stipulations, etc.) and at the pragmatic level of social language use (e.g., information exchange, stance taking or persuasion in debates). Hence, from a linguistic viewpoint, the analysis of infodemic utterances (including fake news, propaganda, troll messages, etc.) focuses primarily on linguistic signals indicating their factuality, claim, opinion, trust, or believability status. Early work on distinguishing liars from truth-tellers tried to identify such cues directly (e.g., liars using fewer self-references and more other-references, avoiding causal terms, using more negative emotion words, and revealing characteristic syntactic patterns) (Burgoon et al. 2003; Newman et al. 2003; Hancock et al. 2007; Lee et al. 2009 ). Yet, a recent survey by Gröndahl & Asokan (2019) provides ample evidence that the search for general stylistic traces of deception, i.e., linguistic markers carrying high emotional load, a high degree of generality/abstractness, high/low use of firstperson pronouns, high use of verbs and certainty-related words, might be doomed to failure.

An alternative avenue of research on testing the believability of assertions does not focus on direct linguistic cues but has its roots in ancient rhetorics (syllogisms) and concentrates on discourse pragmatics. This approach is based on the seminal work of Stephen E. Toulmin (1958) . He introduced the so-called Toulmin schema which distinguishes three fundamental and three auxiliary components of coherent argumentative discourse. According to Toulmin, a claim (thesis, conclusion) can be derived from a piece of information (data) by making use of inference rules (warrants). This standard deductive reasoning scheme can be further complemented by supporting evidence (backing, typically general norms, value sets, moral standards, etc.) adding additional evidence mostly to warrants but also to data. The strength or certainty of information can further be adjusted by modal qualifiers at any stage (such as "mostly'', "probably''), whereas exclusive conditions (rebuttals) can be expressed to indicate exceptions to general rules. Due to its high degree of idealization, the Toulmin schema has stimulated research in many fieldsfrom formal logics and artificial intelligence (Verheij 2009; Caminada 2018) to pragmatics-focused linguistics (van Eemeren and Grootendorst 2004) and computational linguistics (Cabrio and Villata 2012; Hidey et al. 2017) . In particular, the automatic recognition of (im)proper argumentation structure has recently received enormous attention in terms of argumentation mining (Lippi and Torroni 2016; Habernal and Gurevych 2017) . There are also strong links of argumentation process modeling to the fields of software agents (Parsons et al. 1998 ) and multi-agent decision making (Karacapilidis and Papadias 1998) .

Rather than explicitly enumerating specific linguistic cues, current work in the field of automatic natural language processing investigates the potential of automatic classification methods to distinguish between truth-friendly and -unfriendly language use (for a recent survey, cf. Fitzpatrick et al. 2015) , employing deep learning architectures, in particular (Popat et al. 2018; Liu and Wu 2020) . Especially, the areas of fake news detection (Popat et al. 2018; Liu and Wu 2020) and fact/claim checking (Rashkin et al. 2017; Volkova et al. 2017 ; Thorne and Vlachos 2018) have recently attracted a lot of attention.

However, unlike the vast majority of language understanding tasks, (non-expert) persons can have substantial problems in discerning deceptive from non-deceptive language. Actually, their accuracy when it comes to detecting textual deception is approximately on a chance level, or even worse (Bond and DePaulo 2006) so that valid ground truth is hard to attain. Hence, besides testing the performance of deception/fake classifiers the creation of reliable test data sets is a major challenge in current language-focused infodemics research and covers a broad range of topics: Deception and lies (Fitzpatrick and Bachenko 2019), fake news, and fact checking of claims (Augenstein et al. 2019) . They form the benchmarks underlying specialized challenge competitions, e.g., aiming at fact checking in social media (Barrón-Cedeño et al. 2020).

The Cognitive Psychology perspective switches the focus from the way information is designed and disseminated to how pandemics and infodemics are perceived and how beliefs are formed and revised. Research on decision-making and thinking distinguishes between two modes of thinking: One more deliberate and analytic, the other more intuitive and heuristic (Kahneman 2003; Evans and Stanovich 2013) . Intuitive thinking, though often accurate, may lead to systematic biases that appear even more likely in the case of information of low quality or systematic misinformation spreading through social media. For example, people prefer to search for information that supports rather than contests prior beliefs (confirmation bias) and are overly reliant on the believability of conclusions when assessing the quality of arguments (belief bias). Also, the extent to which one's own opinions are shared by others is frequently overestimated (false consensus effect), as is the impression of how widespread beliefs of others are in specific social network structures (majority illusion; Lerman et al. 2016) . Various cognitive tools have been tested in recent years that aim to increase individuals' ability to judge the quality of information and the credibility of sources, and to empower autonomous decisionmaking in general (Lorenz-Spreen et al. 2020).

Despite these organized efforts of misinformation containment and public fact checking, false beliefs tend to persist. Belief updating and revision occurs rather slowly, following the principle of minimal change that does not question the wider belief system or more fundamental core beliefs (Gärdenfors 1992) . Lewandowsky et al. (2012) give an overview of the causal mechanisms which help to explain why people often do not change their beliefs and behavior even if the information on which their beliefs are based has been proven to be false and retracted. When seeking information, people appear to accept the truthfulness of speakers by default and tend to question the quality of information mostly if it is inconsistent with their beliefs, produces an incoherent narrative or overt cues, or if significant others question the credibility of the source. Information is usually part of broader narratives used by everyone to extract meaning from unfolding events. Mere fact checking is more likely to be effective, if it also offers verified alternative narratives for both, the original false and the verified information (Lewandowsky et al. 2012) In a digital environment that develops at a fast pace, risk and digital information literacy are more important than ever. Research shows that risk perception may be distorted if explanations are missing, when actually accurate information is presented in a format that is not well-tuned to the cognitive system of recipients, or when the risks involve low-probability events with particularly threatening consequences (Slovic 2010; Gigerenzer 2015) . For instance, people have difficulties in understanding what it actually means when today's weather report announces a 30% chance of rain, or when a public health organization posts that the relative risk of thrombosis increases by 100% when taking a contraceptive pill. Serious incidents such as the terrorist attacks on September 11th and perhaps COVID-19 incite strong emotions and increase perceived risks way beyond their actual probability. Gigerenzer (2004) showed that in the three months following September 11th, American citizens reduced air travel which in turn led to a substantial increase in fatal car accidents. In order to reduce such hazards, public policy making seems to face the difficult task to not only consider direct, objective risks but also to take factors into account that influence public risk perception and to predict the corresponding behavioral consequences.

In order to increase risk and digital information literacy and autonomous decision-making in general, a number of cognitive tools have been introduced in recent years. Kozyreva, Lewandowsky, and Hertwig (2019) distinguish nudging and boosting. Whereas nudges typically target the design of the choice environment and require little activity from users, boosts address lasting changes in competences of users and require active cooperation. As an example for boosts, the procedure used by professional fact-checkers was translated into training interventions and decision aids that guide users in how to evaluate the credibility of the source, the evidence in arguments and how to read laterally, i.e., cross-check information on other sites (Wineburg and McGrew 2019) . Another type of boost has been termed knowledge-based, deliberate ignorance (Hertwig and Engel 2016) . With an overwhelming amount of easily available information, we routinely decide which information to pick up and which to ignore. Tools supporting deliberate ignorance make use of ratings or other cues to source quality. People seek information to satisfy basic needs such as the need for competence, the need for autonomy and the need to belong (Deci and Ryan 2000) . Both public policies to cope with the infodemic and cognitive tools designed to improve decision-making autonomy, appear most promising the more they consider all three needs together.

From the perspective of Business and Economics, the three views discussed before can be integrated into the wider context of entire economic systems (such as an organization or society). From a macroscopic perspective, such systems can be regarded as complex and adaptive social systems from which some peculiar properties can be abstracted (Thurner et al. 2018 ):

• At the micro-level, such systems consist of a large number of parts which interact with each other in a non-trivial way. For social systems, these parts can, for example, represent individuals, firms, entities which provide, personalize and disseminate information, policymakers, and the government. The interactions between the parts of a complex system are driven by specific laws and channels of interaction. Against the background of infodemics, fundamental channels of interactions are, for example, social media platforms, further means of digital communication and other entities which provide individuals with personalized information. • All micro-elements of such a system have individual characteristics which cover, for example, roles to play in the society, states of information, linguistics patterns in communication, cognitive capacities to form beliefs, the capabilities to fact-check information, ways to revise beliefs, and algorithms to automatically detect misinformation. Together with the laws and channels of interactions, the individual characteristics infer the macro-properties of the entire system. In the context of the COVID-19 pandemic, resilience or specific types of herd behavior could be among such properties at the macroscopic level. From the perspective of Business and Economics, long-term behavioral effects of the COVID-19 pandemic such as travel or buying behavior could be characteristics which are of ultimate interest. • The society adapts to its environment over time, whereby the adaptive change is substantially governed by the system's environment and the information accessible to the parts of the system (Simon 1990; Thurner et al. 2018 ). In the context of hazardous situations such as the COVID-19 pandemic, the system's environment is -in large parts -structured by policy-decisions: Enforceable (behavioral) rules such as compulsory face masks, travel bans, and lockdowns of entire industries shape the limits of a society's room for adaptivity substantially and might govern individual behavior into a specific direction. In addition to the structure of the environment, a complex system's adaptive behavior is also affected by the states of information of its members, which can be interpreted as a function of the information provided to members of a society and their capabilities to collect, decipher, fact-check and make use of this information.

One fundamental question is how societies can be guided in their adaptive behavior so that both the micro-and the macro-properties resemble required characteristics such as preparedness, resilience or a specific type of behavior to mitigate negative effects of the pandemic for organizations or the economy.1 The behavioral dynamics emerging from infodemics might add additional complexity to this question as coordination requires communication, and the efficiency of the communication might be affected by infodemics. .

Infodemics and coordination of individual and collective behavior towards a specific set of objectives are interrelated in a multiplicity of ways.

• First, one may argue that infodemics hinder coordination of behavior towards a specific objective: misinformation might unfold in unwanted behavioral dynamics as it might provide individuals with reasons for not adhering to specific policy decisions or for behaving so that specific objectives cannot be achieved; believers in conspiracy theories, for example, may regard the actions taken by authorities as evidence for their theory and, thus, refuse even more to comply with policies. Through the lens of Business and Economics, such adverse behavior might be of ultimate interest: Understanding the behavioral dynamics which emerge from infodemics are, for example, of high importance in a planning context; understanding the dynamics allows for better anticipating the behavior of an organization's stakeholders, which might help avoiding a substantial amplification of uncertainty in the aftermath of COVID-19.

• Second, certain efforts towards coordination may be made to mitigate infodemics. In the short term, such counter measures could be the design of algorithms for information provision, personalization of provided information, public fact-checking of information or the automated detection of misinformation by the means of computational linguistics. In the long term, coordination efforts could, for example, cover issues of education in order to increase risk-and digital information literacy (e.g., in terms of boosting). In addition to gaining a mere understanding of the dynamics emerging from an infodemic (as discussed above), efficient coordination enables guiding the behavior of a society. In order to do so, it is utterly important to understand the dynamics emerging from infodemics in terms of how individuals and collectives respond to coordination efforts. Policies to strengthen the economy in the aftermath of COVID-19, for example, need to be carefully tested before being launched in order to avoid adverse effects triggered by unprecedented dynamics emerging from the COVID-19 pandemic and the accompanying infodemic. • Third, as coordination involves communication, the question whether communication in the course of coordination could accidentally or deliberately be also part of the infodemic is particularly difficult. Therefore, it is of ultimate importance to understand the dynamics which an infodemic unfolds so that the communication in the course of coordination does not fuel unwanted behavioral, social and economic dynamics but contributes to the guided self-organization of a society so that intended patterns (e.g., in terms of achieved objectives) emerge (Leitner and Wall 2019; Wall 2019).

In this position paper, we particularly focus on the phenomenon of infodemics, i.e., on information and misinformation which spreads rapidly and makes it difficult for decisionmakers to find reliable sources of information. The dynamics emerging from infodemics can be complex and result in adverse individual behavior and might render policies to contain the effects of hazardous events inefficient. Such an infodemic can be observed in the context of the ongoing COVID-19 pandemic. We argue that it is of ultimate importance for the field of Business and Economics to gain a deep and substantial understanding of the dynamics emerging from the current infodemic. We propose to employ powerful techniques like, for example, agent-based modelling which could be a valuable approach for predicting the effects of actions taken in a crisis situation like the COVID-19 pandemic in conjunction with the human behaviour within the crisis (Adam 2020; Squazzoni et al. 2020) Research on infodemics is a strongly interdisciplinary endeavor: It requires the integration of a multiplicity of disciplines. In this position paper, we put a particular focus on the fields of Computer Science and Information Systems, Linguistics, Cognitive Psychology, and Business and Economics. The field of Business and Economics has the potential to integrate the views of these disciplines into a holistic research perspective. We argue that this can be done by taking a complex and self-adaptive system approach: Other disciplines often focus on the level of the individual, for example in terms of individual characteristics such as cognitive capacities, information search behavior or the usage of linguistic patterns; the integration into a complex system allows for inferring the dynamics which emerge from such individual characteristics at the level of an economic system, such as organizations or the society. Understanding these dynamics is of ultimate importance for the field of Business and Economics, as it supports organizations and policy-makers in anticipating and coordinating behavior and, thereby, helps reduce the amplification of uncertainty potentially resulting from the COVID-19 pandemic.

This position paper focuses on the fields of Computer Science and Information Systems, Linguistics, Cognitive Psychology, and Business and Economics. In order to get a full and deep understanding of the dynamics emerging from infodemics, the integration of further scientific fields, such as Sociology, Ethics, and the Legal Sciences is strongly needed. Thus, aside from highlighting the importance of the considered fields for research on infodemics, we conclude with a call to action for strongly interdisciplinary research on this topic. Such exceptional times require us to look beyond the horizon of our own discipline and to join forces in order to unfold collective creativity.

Multistakeholder recommendation: Survey and research directions

Modelling the pandemic: The simulations driving the world's response to COVID-19

MultiFC: A real-world multi-domain dataset for evidence-based fact checking of claims

Trump's aggressive advocacy of malaria drug for treating coronavirus divides medical community

Covid-induced economic uncertainty

Collective Action

CheckThat! at CLEF 2020: Enabling the Automatic Identification and Verification of Claims in Social Media

Accuracy of deception judgments

Detecting deception through linguistic analysis

Multisided fairness for recommendation

Combining textual entailment and argumentation theory for supporting online debates interactions

Rationality postulates: applying argumentation theory for non-monotonic reasoning

Controlling polarization in personalization: An algorithmic framework

The covid-19 social media infodemic

The "what" and "why" of goal pursuits: Human needs and the selfdetermination of behavior

Dual-process theories of higher cognition: Advancing the debate

Economic effects of coronavirus outbreak (COVID-19) on the world economy

Building a forensic corpus to test language-based indicators of deception

Blockbuster culture's next rise or fall: The impact of recommender systems on sales diversity

On the (im)possibility of fairness

Belief revision: An introduction

Dread risk, September 11, and fatal traffic accidents

COVID-19 and finance: Agendas for future research

Text analysis in adversarial settings: Does deception leave a stylistic trace

Argumentation mining in user-generated web discourse

On lying and being lied to: A linguistic analysis of deception in computer-mediated communication

Homo ignorans: Deliberately choosing not to know

Analyzing the semantic types of claims and premises in an online persuasive forum

A perspective on judgment and choice: mapping bounded rationality

Hermes: Supporting argumentative discourse in multi agent decision making

Features of computer-mediated, text-based messages that support automatable, linguistics-based indicators for deception detection

On the dynamics emerging from pandemics and infodemics

Decision-facilitating information in hidden-action setups: An agentbased approach

Covid-19: US gives emergency approval to hydroxychloroquine despite lack of evidence

The" majority illusion" in social networks

Misinformation and its correction: Continued influence and successful debiasing

Argumentation mining: State of the art and emerging trends

FNED: A deep network for fake news early detection on social media

Hertwig R (2020) How behavioural sciences can promote truth, autonomy and democratic discourse online

Lying words: Predicting deception from linguistic styles

The socio-economic implications of the coronavirus and COVID-19 pandemic: a review

DeClarE: Debunking fake news and false claims using evidence-aware deep learning

COVID-19 infodemic: More retweets for science-based information on coronavirus than for false information

Truth of varying shades: Analyzing language in fake news and political fact-checking

Covid-19-related web search behaviors and infodemic attitudes in italy: Infodemiological study

Invariants of human behavior

Computational Models That Matter During a Global Pandemic Outbreak: A Call to Action

The uses of argument

Automated fact checking: Task formulations, methods and future directions

Introduction to the theory of complex systems

Drug (2020) FDA cautions against use of hydroxychloroquine or chloroquine for COVID-19 outside the hospital setting or a clinical trial due to risk of heart rhythm problems

Infodemic and risk communication in the era of CoV-19

A systematic theory of argumentation: The pragma-dialectical approach

Separating facts from fiction: Linguistic models to classify suspicious and trusted news posts on twitter

Coordination with erroneous communication: results of an agent-based simulation

Lateral reading and the nature of expertise: Reading less and learning more when evaluating digital information

Managing epidemics: key facts about major deadly diseases. World Health Organization