Tutorial Summary WSDM ’19, February 11–15, 2019, Melbourne, VIC, Australia

Fake News: Fundamental Theories, Detection Strategies and

Challenges
Xinyi Zhou, Reza Zafarani Kai Shu, Huan Liu
Data Lab, EECS Department Computer Science and Engineering
Syracuse University Arizona State University
{zhouxinyi,reza}@data.syr.edu {kai.shu,huan.liu}@asu.edu

ABSTRACT 1 CONTEXT AND MOTIVATION

The explosive growth of fake news and its erosion to democracy, jus- Fake news is now viewed as one of the greatest threats to democracy,
tice, and public trust increased the demand for fake news detection. journalism, and economies. It has weakened public trust in govern-
As an interdisciplinary topic, the study of fake news encourages a ments and its potential impact on the contentious “Brexit” referen-
concerted effort of experts in computer and information science, dum and the equally divisive 2016 U.S. presidential election - which
political science, journalism, social science, psychology, and eco- it might have affected [1] - is yet to be realized. The reach of fake
nomics. A comprehensive framework to systematically understand news was best highlighted during the critical months of the 2016
and detect fake news is necessary to attract and unite researchers U.S. presidential election campaign, where top twenty frequently-
in related areas to conduct research on fake news. This tutorial discussed false election stories generated 8,711,000 shares, reactions,
aims to clearly present (1) fake news research, its challenges, and and comments on Facebook, ironically, larger than the total of
research directions; (2) a comparison between fake news and other 7,367,000 for the top twenty most-discussed election stories posted
related concepts (e.g., rumours); (3) the fundamental theories de- by 19 major news websites [12]. Our economies are not immune to
veloped across various disciplines that facilitate interdisciplinary fake news either, impacting stock markets and leading to massive
research; (4) various detection strategies unified under a comprehen- trades. For example, fake news claiming that Barack Obama was
sive framework for fake news detection; and (5) the state-of-the-art injured in an explosion wiped out $130 billion in stock value [8].
datasets, patterns, and models. We present fake news detection The generous benefits in fake news activities are one of the moti-
from various perspectives, which involve news content and infor- vations for people to initiate and engage in such activities. Consider
mation in social networks, and broadly adopt techniques in data dozens of “well-known” teenagers in the Macedonian town of Veles
mining, machine learning, natural language processing, informa- who produced fake news for millions on social media and became
tion retrieval and social search. Facing the upcoming 2020 U.S. pres- wealthy by penny-per-click advertising during the 2016 U.S. presi-
idential election, challenges for automatic, effective and efficient dential election [13]. Such stories attach greater importance to fake
fake news detection are also clarified in this tutorial. news detection and intervention as they provide an incentive for
individuals to become the next “Macedonian teenagers” in the up-
CCS CONCEPTS coming 2020 U.S. presidential election. On the other hand, with fake
news detection research in its early stages, greater opportunities
• Human-centered computing → Collaborative and social com-
exist for malicious individuals to create and spread fake news in
puting theory, concepts and paradigms; Empirical studies in collabo-
the absence of a worry. To facilitate further development in this
rative and social computing; • Social and professional topics →
area, we address both theoretical and technical aspects of fake news
Computer crime; • Applied computing → Computer forensics;
detection in this tutorial:

KEYWORDS (1) Interdisciplinary fake news research is encouraged, where fund-

Fake news; fake news detection; news verification amental theories developed across disciplines can facilitate qual-
itative and quantitative studies, as well as developing well-
ACM Reference Format: justified and explainable fake news detection techniques.
Xinyi Zhou, Reza Zafarani and Kai Shu, Huan Liu. 2019. Fake News: Fun- (2) A comprehensive framework and strategy to systematically
damental Theories, Detection Strategies and Challenges. In The Twelfth
understand and detect fake news is necessary. Such strategies
ACM International Conference on Web Search and Data Mining (WSDM ’19),
attract and unite researchers mastering knowledge and tech-
February 11–15, 2019, Melbourne, VIC, Australia. ACM, New York, NY, USA,
2 pages. https://doi.org/10.1145/3289600.3291382 nologies in related areas to work on fake news topic.
(3) Open issues and challenges for fake news studies should be
clarified to highlight future research directions and priorities.
Permission to make digital or hard copies of part or all of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation 2 TUTORIAL OUTLINE
on the first page. Copyrights for third-party components of this work must be honored.
For all other uses, contact the owner/author(s). Preventing fake news from impacting democracy, journalism and
WSDM ’19, February 11–15, 2019, Melbourne, VIC, Australia economies demands researchers, practitioners, and industry leaders
© 2019 Copyright held by the owner/author(s). to attach great importance to understanding and detecting fake
ACM ISBN 978-1-4503-5940-5/19/02.
https://doi.org/10.1145/3289600.3291382 news. Given a clear definition of fake news, this tutorial presents a

836
Tutorial Summary WSDM ’19, February 11–15, 2019, Melbourne, VIC, Australia

comprehensive survey of fake news research. In particular, the tuto- brings about new challenges. The tutorial presents the open issues
rial (1) identifies fundamental theories across various disciplines; (2) that are important but have not been (well) addressed in recent
elaborates the detection strategies under a comprehensive frame- studies. It points out the potential resources (e.g., fact-checking web-
work and further introduces the related datasets, patterns, models, sites) and techniques (e.g., deep learning) that are able to address
and algorithms; (3) clarifies the open issues in the state-of-the-art, the open issues and challenges. The tutorial also highlights several
and challenges to be faced for the development of fake news studies. tasks as future research directions, which can improve the perfor-
Fundamental Theories. Human vulnerability to fake news, which mance of fake news detection, and promote our understanding of
can bring in useful clues or further complicate fake news detec- fake news (e.g., identifying check-worthy content).
tion, has been a subject of interdisciplinary research. For instance,
achievements in forensic psychology such as Undeutsch hypothe- 2.1 Target Audience and Prerequisites
sis [14] have pointed out the style difference between truth and de- The tutorial would be interesting for researchers, students, practi-
ceptive information. Similarly, interdisciplinary research has looked tioners, and project managers in areas such as Computer Science
at why individuals spread fake information, considering that the and Engineering, Information Science and Management, Journal-
borderline between malicious and normal users becomes unclear ism, Political Science, Social Sciences, Psychology and Economics.
– normal people can frequently and unintentionally participate Preliminary background in data mining, machine learning, natural
in fake news activities as well, due to, e.g., social identity [2] or language processing is recommended for tutorial participants.
self-preexisting knowledge [6]. This tutorial conducts a compre-
hensive literature study across various disciplines. We review more 2.2 Resources
than twenty well-known theories that can contribute to our un- The tutorial summarizes current state of fake news research. In
derstanding of fake news and participants in fake news activities. particular, the tutorial has a companion survey paper [16]. Other
We present and discuss the problems arising as explained by these resources we recommend are two overview papers [9, 11], a policy
theories, ranging from the patterns they can reveal, the qualitative forum [5], and a related tutorial [15].
and quantitative fake news studies one can conduct based on these Resources Availability. The videos, slides, related papers, datasets
studies, to the specific roles they can play to detect fake news. and tools are all available and timely updated at the following web-
Detection Strategies. Detecting fake news is a complex and mul- site: https://www.fake-news-tutorial.com/.
tidimensional task due to the characteristics of fake news. The
detection strategies exploit multiple news-related (e.g., headline, REFERENCES
body text, publisher) and social-related (e.g., feedback, propagation [1] Hunt Allcott and Matthew Gentzkow. 2017. Social media and fake news in the
paths and spreaders) types of information. Each information type 2016 election. Journal of Economic Perspectives 31, 2 (2017), 211–36.
[2] Blake E Ashforth and Fred Mael. 1989. Social identity theory and the organization.
can be in the form of text, multimedia, network, etc., corresponding Academy of management review 14, 1 (1989), 20–39.
to various applicable techniques and usable resources. The tutorial [3] Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Mur-
phy, Thomas Strohmann, Shaohua Sun, and Wei Zhang. 2014. Knowledge vault:
reviews the detection of fake news from four perspectives of knowl- A web-scale approach to probabilistic knowledge fusion. In Proceedings of the 20th
edge, style, propagation and credibility. Specifically, from a knowl- ACM SIGKDD international conference on Knowledge discovery and data mining.
edge perspective, fake news detection is a “comparison” between ACM, 601–610.
[4] Nitin Jindal and Bing Liu. 2008. Opinion spam and analysis. In Proceedings of the
the relational knowledge extracted from the to-be-verified news 2008 International Conference on Web Search and Data Mining. ACM, 219–230.
articles and that of knowledge-bases representing facts/ground [5] David MJ Lazer, Matthew A Baum, Yochai Benkler, Adam J Berinsky, Kelly M
truth [7]. Style-based fake news detection aims to capture and Greenhill, Filippo Menczer, Miriam J Metzger, Brendan Nyhan, Gordon Penny-
cook, David Rothschild, et al. 2018. The science of fake news. Science 359, 6380
quantify the differences in writing styles between fake and true (2018), 1094–1096.
news. Propagation-based fake news detection uses information pro- [6] Raymond S Nickerson. 1998. Confirmation bias: A ubiquitous phenomenon in
many guises. Review of general psychology 2, 2 (1998), 175.
vided in news dissemination. Finally, credibility-based fake news [7] Jay Pujara and Sameer Singh. 2018. Mining Knowledge Graphs From Text. In
detection assesses the credibility of headlines (e.g., using click-bait Proceedings of the Eleventh ACM International Conference on Web Search and Data
detection [11]), publishers (i.e., source websites), comments (e.g., Mining. ACM, 789–790.
[8] K Rapoza. 2017. Can ‘fake news’ impact the stock market?
using opinion spam detection [4]), and users to indirectly detect [9] Kai Shu, H Russell Bernard, and Huan Liu. 2018. Studying Fake News via Network
fake news. Each perspective carries its own usable set of tools [3], Analysis: Detection and Mitigation. arXiv preprint arXiv:1804.10233 (2018).
datasets [10] and various detection strategies in data mining, ma- [10] Kai Shu, Deepak Mahudeswaran, Suhang Wang, Dongwon Lee, and Huan Liu.
2018. FakeNewsNet: A Data Repository with News Content, Social Context and
chine learning, natural language processing, information retrieval Dynamic Information for Studying Fake News on Social Media. arXiv preprint
and social search. Various perspectives can be integrated under arXiv:1809.01286 (2018).
[11] Kai Shu, Amy Sliva, Suhang Wang, Jiliang Tang, and Huan Liu. 2017. Fake news
a unified framework for fake news analysis, which looks at fake detection on social media: A data mining perspective. ACM SIGKDD Explorations
news from the time being created and published to the time being Newsletter 19, 1 (2017), 22–36.
disseminated. We review, summarize, compare and evaluate current [12] Craig Silverman. 2016. This analysis shows how viral fake election news stories
outperformed real news on Facebook. BuzzFeed News 16 (2016).
studies within this framework during the tutorial. [13] Alexander Smith and Vladimir Banic. 2016. Fake News: How a partying Macedo-
Challenges. News characteristics such as timeliness and oddity1 nian teen earns thousands publishing lies. NBC News 9 (2016).
[14] Udo Undeutsch. 1967. Beurteilung der glaubhaftigkeit von aussagen. Handbuch
indicate that the detection of fake news does not follow that of other der psychologie 11 (1967), 26–181.
fake information, e.g., fake statements and fake reviews, and thus [15] Liang Wu, Fred Morstatter, Xia Hu, and Huan Liu. 2016. Mining misinformation
in social media. Big Data in Complex and Social Networks (2016), 123–152.
[16] Xinyi Zhou and Reza Zafarani. 2018. Fake News: A Survey of Research, Detection
1 https://www.axiapr.com/blog/elements-of-news Methods, and Opportunities. arXiv preprint arXiv:2492706 (2018).

837