Facebook is one of the largest and most widely used social media platforms in the world. With over 2 billion monthly active users, it has changed how people connect and share information globally. This growth has also raised questions about how information spreads on the platform and what impact it has. Researchers have conducted many studies looking at the flow of information on Facebook through network analysis and other techniques.
Some key findings from this research include:
Strong homophily exists in how information spreads on Facebook. Homophily refers to the principle that connection between similar people occurs at a higher rate than among dissimilar people. Studies have found users are much more likely to share content from friends who are similar to them in terms of age, location, education level, and interests. This reinforcement of ideas among homogeneous groups can potentially lead to increased polarization.
Network structure plays an important role in virality. Content that originates in densely connected clusters of the network is more likely to go viral than content randomly posted. Dense clusters allow information to cascade more quickly through the network via confirmation bias, social influence, and increased visibility. Some studies have also found isolated super-spreaders whose reach is disproportionate to their network size.
Emotional content is more likely to spread. Posts containing emotionally charged content like anger, anxiety or awe tend to be shared more on social media compared to neutral information. Positive emotions like joy also facilitate more transmission. This likely relates to evolutionary instincts to affiliate with others during emotional arousal and threats. Highly emotional content might spread at the cost of nuanced discussion.
Popularity bias drives engagement. News stories and websites shared by a higher number of friends receive more likes and comments from a user’s network compared to less widely shared ones. This relates to the importance of social proof in human decision making where people are more inclined to socially endorse popular opinions. Early momentum also predicts how viral a piece of content might become.
Echo chambers and filter bubbles exist. Through their friend networks and algorithmic newsfeeds, users end up primarily exposed to opinion-reinforcing content. Studies have found their sharing behavior also becomes more ideologically extreme over time as they interact mainly within like-minded groups. This exacerbates political polarization via confirmation bias and out-group hostility. The existence and degree of echo chambers is still debated.
Cross-cutting exposure is limited. While users continue engaging within homogeneous clusters, they have relatively infrequent exposure to ideologically different viewpoints due to social contacts being overwhelmingly similar. This limits discussion across lines of difference and potential for consideration of opposing views. Some evidence points to users insulating themselves from opposing opinions too.
Bot activity strongly impacts virality. Twitter and Facebook networks analyzed by researchers show bot-driven amplifications play an outsized role in the spread of political hashtags and links. Bot tweets receives much higher engagement ratios compared to human tweets. Coordinated bot networks are thus able to artificially manufacture trends and influence public narratives by gaming social algorithms.
Facebook algorithms produce unintended consequences. Changes to the EdgeRank algorithm that determines News Feed content have significantly impacted what users see. Tweaks like prioritizing content from friends led to the rise of hyper-partisan media outlets and financially motivated “hyper-sharers.” Clickbait and false news also spread more due to such algorithmic biases emphasizing engagements. Fact-checking and changes to increase contextual integrity have met limited success.
Foreign interference campaigns effectively targeted users via ads and pages. Russian troll operations and political ads were targeted along geographic, ideological and culturally salient divisions to exacerbate societal tensions during elections. User data harvested by third-party apps enabled extremely precise micro-targeting outside traditional media markets. Lack of transparency around political ads and inauthentic behavior remain problems.
Social evidence products spread misinformation inadvertently. Features like Trending Topics and Facebook Video autoplay sometimes amplified low-credibility content by presenting it without proper context or fact-checking. The automated, algorithmic nature of these curation tools allows falsehoods to amass social proof before being corrected, misleading unwitting users drawn by popularity cues in their feeds.
Research into information flow on Facebook highlights both the platform’s positive affordances for connection while also revealing mechanisms that enable the spread of misinformation at unprecedented scales with real-world impacts. Identifying pathways for the circulation of inflammatory, intentionally deceptive and conspiratorial content through things such as emotional manipulation, bot networks or foreign interference campaigns has major implications to address the threat of “fake news.” At the same time, changes to optimize the algorithms or transparency around political advertising and authenticity are difficult to balance against user experience and business priorities. More work is still required to understand the dynamics of online influence campaigns amidst complex human social behavior mediated by technology design decisions underlying such platforms. Future studies must continue exploring ethical ways to shape platform affordances and user behavior to enhance informed public discourse online.
Andrew Stroud