Panel 2: “Fact Checkıng, Fake News and Bot Detectıon”

Panel 2: “Fact Checkıng, Fake News and Bot Detectıon”

Banu Akdenizli

Media in the Middle East: Navigating the Digital Landscape Amid Fake News and Propaganda

Increased activity in the Twittersphere from government officials and the public; concern over fake news and how to combat it in the region; government surveillance and data gathering; privacy and VPNs. These are just some of the issues making cybersecurity a new geopolitical challenge in the GCC region. Arab audiences are heavy online news consumers. In fact, behind socializing, consuming news is the next most popular online activity. Politics and current affairs are among the most popular types of news (Dennis, Martin & Hassan, 2019). Despite their heavy consumption, Arab Internet users were aware of the dangers of fake news. In a study completed in 2017; 43%say they were very concerned about the spread of false news online, and 33% expressed concern (Salem, 2017). Where is the region now? Despite regional differences, it is unsurprising that internet users’ concerns about privacy, surveillance, and fake news track similarly on a global scale given the nature of technology and the increasing threats to users. GCC governments continue to feel threatened about social media use, evident in the legislation and severe penalties concerning social media. With a particular focus on Qatar, this presentation will consider the latest statistics from the region, current legislation, and efforts to combat disinformation.

Hamdi Akın Ünver

Can Fact-Checking Combat Computational Propaganda? Evidence from Turkey

How can we combat computational propaganda? Should governments battle it, should platforms be entrusted with this mission, or should we employ fact-checkers? This paper follows the third-route and tests the performance of Turkish fact-checkers against disinformation and computational propaganda. It does so by conducting a temporal analysis of the two most popular fact-checking platforms and tests their performance in March 2019 elections via a dataset of 74 million tweets. Ultimately, the study finds that fact-checkers’ success is very context-specific and is usually determined by the quality (or lack thereof) of computational propaganda tactics themselves. In addition, the paper ends with a nationally-representative survey demonstrating how Turks view fact-checkers and whether they use fact-checking as part of their information-seeking behaviour.

Marc Owen Jones

Tracking Adversaries: The Evolution of Manipulation Tactics on Gulf Twitter

Despite well  publicised attempts at removing Computational Propaganda on Twitter, it is still widespread on Persian Gulf Twitter. This paper documents the prevalence of the frequent engagement of sockpuppet accounts on hashtags connected to Saudi Arabia. In particular, it explores the continued innovation in tactics by sockpoppet operators in circumventing automatic spam detection. In particular it explores the evolution of sock puppets and bots since 2016, noting how while Twitter appears to be better at purging accounts based on user creation date, this is increasingly less common, and other network markers such as centrality are a better marker for sockpuppet accounts. By tracking the evolution of known sock puppet and bot operators since 2015 using network analysis,  this paper charts the evolution of spam circumvention, but also raises attended issue of governance, for example? Why does such easily detectable manipulation remain in play for so long. Given revelations by a former Facebook employee that content moderation is dealt with according to political expediency, this paper posits whether Middle East Twitter is considered a market ripe for exploitation, but not governance. 

Onur Varol

In this talk, I will introduce the social bot detection framework Botometer and its applications for investigating social bot activity in political conversations, fake engagement, and purchased followers. I will discuss our efforts to keep current models up-to-date by introducing machine learning architectures specialized for different bot behaviors. Recent manipulation activities use sophisticated techniques to influence crowds, so I will highlight the need to analyze bot activities from a coordination perspective.