Carnegie Mellon University

Center for Informed Democracy & Social - cybersecurity (IDeaS)

CMU's center for the study of disinformation, hate speech and extremism online

IDeaS Center for Informed Democracy & Social-cybersecurity

Photo free to use from Unsplash.

January 17, 2024

The Role of Narratives and Networks in the Spread of Online Hate

By Joshua Uyheng

Photo free to use from Unsplash.

Publication Citation

Uyheng, J., & Carley, K. M. (2023). Online hate in the Philippines: The role of narratives and networks. International Review of Law, Computers & Technology. Advance online publication.

Why does hatred spread online? Despite its harmful reputational effects and the social norms which discourage it, online hate proliferates like a disease and infects our everyday conversations and relationships with others.

This is particularly true in times of crisis and societal upheaval. In moments when people arguably need to come together the most, hatred seems to come out in full force to tear people apart. And rather than encourage more and deeper connections, social media has played an important role in accelerating such divisive processes.

In a recent publication, we sought to examine two key social mechanisms that might help explain the spread of online hate: narratives and networks. Using the Philippines as a study site, we investigated online hate in the context of a recent national election and the initial outbreak of the Covid-19 pandemic. Interoperable computational social science tools were designed and deployed to characterize online hate narratives and networks in these settings, as well as assess their potential manipulation at the hands of bot-driven information operations.

Hate Narratives

Using computational text analytical techniques, we derived focal themes in the hate narratives which attended online conversations surrounding the elections and the pandemic in the Philippines.

During the elections, ­­­political and religious narratives were blended to mobilize online hate across political camps. Many hateful statements were made by simply using derogatory language in conjunction with the mainstream political slogans used by the electoral candidates. Yet in a predominantly Roman Catholic country like the Philippines, religious conflicts frequently intertwine with political contests, and the recent election was no exception. For both sides, supporters of opposing camps were framed as morally hypocritical for allying with corrupt politicians.

Meanwhile, when the pandemic first broke out, we discovered that the central narratives of online hate had to do with race. Racial narratives abounded decrying the implicit hierarchy of Filipinos and Chinese people in the country. Those expressing hateful sentiments believed that Filipino lives were being valued less than those of Chinese foreigners. This claim, in turn, was anchored in the political posturing of then-President Duterte to curry favor with China. Because of this belief, online hate was directed toward both the government and to the Chinese people whose welfare was deemed to have been prioritized over the Filipino people’s in a time of crisis.

Across both cases, hate narratives drew upon a sense of injustice committed by the targeted groups against the speakers’ own communities. In utilizing such framing devices, the expression of online hate could be made to appear legitimized and aided in its proliferation through social media as justified attacks against wrongdoing. Narratives thus act as a powerful catalyst for the spread of online hate.

Hate Networks

While the content of online hate uses narratives to put various concepts and meanings together, the people which spread online hate use networks to coordinate and amplify their hateful messages. Interestingly, notable differences were detected between the hate networks at play in the elections and the pandemic.

Common across both events was the fact the online hate networks sought to infiltrate the mainstream. This was in stark contrast to our previous findings which saw that online hate networks generally benefited from isolation from the mainstream. Due to the acute nature of the events, it may have been the case that hate groups wanted to influence mainstream discourse at large rather than sequester and radicalize in more typical echo chamber structures.

A crucial difference between the two events, however, was that hate groups during the elections were much more organized, featuring higher levels of hierarchy and density than those during the pandemic. This may likely have been due to the unexpected and volatile nature of the pandemic, which arose unexpectedly. In contrast, the elections had long been anticipated to become a flashpoint of societal divisions in the Philippines, and may likely have launched hateful sentiments from more well-established hostile camps.

Manipulating Narratives and Networks

Finally, this research used narratives and networks as a lens to assess the potential manipulation of online hate. Using the BotHunter tool, we identified which accounts participating in the online conversation were likely to be automated accounts. From there, we could then assess their involvement in the spread of particular hate narratives and the roles they played in hate networks.

Across both events, it was noteworthy to observe that bots inserted themselves into expressions of online hate with a similar narrative signature. For both elections and the pandemic, bots were most likely found utilizing hate narratives that drew on political and religious identities, while straying away from racial and gender narratives. Bots thus seemed to stoke existing political hostilities in the country, but it was humans that were more likely to express anti-Chinese sentiments during the pandemic.

Meanwhile, along the network dimension, clear distinctions were again detected. In the elections, the presence of bots in hate networks predicted increased organization. However, during the pandemic, bots were more likely to appear in less dense communities. During the Covid-19 crisis, then, it seemed that bots sought to add fuel to the fires of toxic online discourse, but not in a particularly orchestrated manner. Conversely, during the election, they were highly organized and may have been used as a well-coordinated tool to amplify online hate during the campaign.


Around the world, online hate poses a serious problem for social media platforms and society at large. Particularly in high-stakes contexts like pandemics and elections, online hate can threaten social cohesion and widen societal divides. Through a narrative and network approach, we can better understand how online hate accomplishes these harmful functions. Moreover, narratives and networks help us better understand how online hate can also be manipulated at these crucial moments, thus informing more robust and resilient responses to such threats.