Shelby Grossman, Katie Jonsson, Nicholas Lyon, and Lydia Sizer

In this study, we use Libya as a case to assess how social media content about a polarizing, conflict-related event varies by the country of the information producer. We created a dataset of Facebook posts about a sHaftar’s attack on Tripoli in 2019.


Foreign social media outlets with political interests can use social media platforms to attempt to influence a country’s politics. In fragile contexts such as Libya, where both social media penetration and information uncertainty are high, these efforts could be especially impacteful.

We find that more than half of the posts originated from outside Libya and that posts from countries aligned with the former Tripoli-based government are biased towards it and posts from countries aligned with the eastern-based Haftar are biased toward his forces. However, many Pages are not slanted: the correlations are instead driven by a smaller number of hyper-partisan Pages.

Our findings have implications for our understanding of how social media content – especially from abroad – could shape citizen perceptions of the legitimacy of competing political actors.


People are increasingly relying on social media for news. Thirty-six percent of American adults younger than 30 report often getting news from social media. In Argentina, another country for which there is reliable data on this topic, 78% of people who follow the news use social media as a news source.

This reliance is potentially worrisome given the lack of transparency on many social media accounts: anyone in any country can create a Facebook Page or Twitter account and claim to be a credible news outlet. Social media users may not know who created the content on their feeds and may trust accounts that resemble news providers.

This is troubling – especially in uncertain, conflict-riven contexts – as foreign social media content could be part of an overt or covert influence operation.

To what extent does social media content about a country originate outside of that country, and does foreign content differ from domestic content?

We assess these questions by investigating the content and origin of public posts on Facebook Pages about Libya in April 2019. During that month, eastern-based Khalifa Haftar and his self-styled Libyan Arab Armed Forces (LAAF, also known as the Libyan National Army (LNA)) attempted to seize Tripoli, the capital, from the UN-backed Government of National Accord (GNA) headed by then Prime Minister Fayez al-Sarraj.

The military battle for Tripoli played out as an internationalized civil war, with Egypt, France, Russia, Saudi Arabia, and the UAE engaging in unconventional warfare in support of the LAAF, and Turkey and Qatar supporting the GNA through similar tactics.

Using a Facebook-owned product called CrowdTangle, we created a data set of Facebook Page posts, in Arabic and English, about the start of Haftar’s Tripoli offensive in April 2019.

In the month that followed Haftar’s offensive, there was a flood of Facebook posts describing the event from media outlets of varying legitimacy. Some framed the story as an internationally recognized government defending itself from a warlord, while others framed it as a general bringing security to Libya.

For each post, we manually coded the locations of the Page administrators using Facebook’s Page Transparency feature: Facebook’s best guess at the administrators’ locations, making an assessment based on all of the data at their disposal.

The data are useful particularly in cases where administrators attempt to use their self-declared location to conceal their real location. We then compiled a dictionary of pro-LAAF terms and one of pro-GNA terms to create a GNA slant measure for all posts and looked at the relationship between slant – the percent of words in a post aligned with the GNA or LAAF –and Page administrator location.

In total, our data set includes 16,662 posts (89% in Arabic, 11% in English).

We find that more than half of the posts about the Tripoli offensive were from Pages with a plurality of administrators based outside of Libya, and 10% of posts have a plurality of administrators who have opted into hiding their location.

We also find that there is a substantively significant relationship between the location of the content producers and the slant of the post: posts from countries aligned with the GNA (Turkey and Qatar) are more pro-GNA and posts from countries aligned with the LAAF (the UAE and Egypt) are more pro-LAAF.

Yet many Pages are not slanted; these correlations are instead driven by a subset of highly biased Pages. To better understand this surprising finding, we dig into the Pages producing the most slanted coverage and find that an over whelming majority on both sides have ties to foreign countries.

Overall, these findings suggest that during this important month, Facebook content about Libya was heavily influenced by foreign actors and that the war in Libya is fought fiercely in non-kinetic domains by state and non-state actors.

Foreign social media influence operations can be especially destabilizing in places where foreign countries meddle in domestic politics – countries like Syria, Sudan, Venezuela, Mozambique, and Libya.

Since the early 2000s, social media has evolved from a tool used by weaker actors in the international system to set agendas and compel stronger actors to change behavior, to a tool used by stronger states to manipulate regional and international dynamics without overt use of force>

Information warfare on social media can be especially attractive to states due to the plausible deniability if the operative is uncovered and the ability to make content appear as if it originated locally.

Dozens of social media platform public take downs of state-backed disinformation operations show that these information operations are prolific and target countries where other forms of foreign meddling proliferate. We note, however, that many of the foreign social media posts about Libya are not part of an influence operation.

Libya is a useful case for several reasons:

First, there is evidence that Egypt, Russia, Saudi Arabia, and the UAE have conducted online disinformation campaigns about Libya on Twitter and Facebook.

Second, at the time the research was conducted, both al-Sarraj and Haftar struggled for legitimacy, and we can see how online media lowers the barriers for each side to attempt to delegitimize the other.

Third, while researchers and reporters pay much attention to how foreign actors use social media to influence citizens in high-profile cases such as the US, Hong Kong, and the UK, there has been less attention to the vulnerabilities of developing countries.

Fourth, Libya is a useful case because of new developments to its media space: social media flourished after Qadhafi fell. In 2019, 31% of Libyans said their primary source of information was Facebook. And fifth, research suggests that information operations can be more effective when people are less certain about the truth; because conflict creates high degrees of information uncertainty, Libya shows the potential consequences of these campaigns.

This study contributes to three literatures:

First, we contribute to research on social media and polarization. By conducting a deep dive into social media content around a highly salient and polarizing conflict, where the content producers were both domestic and foreign, we can shed light on why social media may have such effects.

We show that the content of social media posts is itself polarized, indicating that content producers can frame the same events indissimilar and slanted ways.

Second, reasearchers have introduced the concept of “sharp power” to describe authoritarian information warfare. Whereas soft power aims to win “hearts and minds”, sharp power is used to manipulate peoples’ information environment.

This concept emerged to capture an increasingly important feature of influence campaigns, but we believe we are the first to attempt to measure the potential scale of sharp power operations in a particular context.

Third, there is an important literature on consumption of false stories.

Increasingly, however, disinformation – defined as the intentional creation and sharing of information with the intent to deceive – takes the form of unfalsifiable hyperpartisan content, spread by fake accounts.

In this paper we show that the foreign backers of competing political actorsleverage unfalsifiable content to shape perceptions of the legitimacy of the opposing side.

Clashing narratives: A theory of slanted social media in conflict settings

There is a large literature suggesting social media and online media can politically polarize citizens – though perhaps no morethan offline media – and shape political attitudes and behavior, including vote choice and protest.

The forces behind the production of politically polarized media, however, have received less attention.

Who is creating the polarizing content and what motivates them to create it?

In a domain that seems to collapse physical space, what is the role of geopolitics in the creation of polarizing content?

In this section we first introduce a theory for why foreign actors willfind social media information operations appealing in countries experiencing civil war.

Second,we theorize how the location of the social media content producer relates to the slant of the content about countries experiencing civil war.

Why foreign actors run information operations on social media

Countries in the midst of civil war are some of the most important contexts for understanding social media and political polarization, as civil war creates high levels of information uncertainty. Access to traditional media may be disrupted, increasing reliance on social media forinformation.

Individuals in these contexts may be more persuadable; research shows individuals will be more likely to update their beliefs in response to new information when they are more uncertain.

Supplying information in a highly uncertain environment is one way of influencing citizen beliefs about the status of the conflict. Moreover, a large literature shows that media can shape individuals’ allegiances and behaviors.

There are several incentives for factions involved in civil unrest, and especially their external patrons, to exploit these tools, especially social media. First, social media gives a va-riety of actors, including foreign governments, domestic political actors, domestic citizens, and international bodies, insight into public opinion about a policy or individual.

Social media is particularly likely to serve this role during civil war, when there are greater obstacles to running public opinion surveys. This can incentivize foreign actors to manipulate the perception of public opinion, for example by engaging in astroturfing.

This refers to adeceptive campaign to inflate the appearance of support for an issue or individual. The term astroturfing comes from the notion of deception: the support is not grassroots, but rather akin to astroturf.

These campaigns make it appear as if accounts are ordinary citizens speaking ontheir own volition. By leveraging the social influence of perceived peers, astroturfing can be used in attempts to shift allegiances in conflicts, much as peer influence has been shown to work in other domains.

Empirically, many known disinformation campaigns – for example, networks suspended and publicized by Facebook and Twitter –have engaged in astroturfing, including in operations trying to sway public opinion about the popularity of particular political actors during civil war.

This raises a second attractive aspect of social media for foreign actors: plausible deniability.

To continue in PART (II)


Shelby Grossman and Katie Jonsson with Stanford Internet Observatory. Nicholas Lyon with Stanford University and Lydia Sizer Tufts University.


Source: Journal of Quantitative Description: Digital Media (2022), 1–57

Related Articles