Skip to Main Content

← Back to Resources

Coronavirus Goes Viral: A Look at How Yonder Responds When The News Happens


While the rapid spread of Covid19 (popularly known as Coronavirus) continues to capture headlines and confound health experts, we here at Yonder are busy studying the spread of narratives about the virus.

Our platform, Yonder Narrative, picked up the obvious uptick in online chatter analyzing 15 million public posts across social media platforms, including 4chan and Gab, from January 21 to February 12, 2020.

But how much of that conversation is authentic and indicates genuine consensus or interest? Click below to watch a webinar hosted by PRNews where we discuss how conversations about the virus circulate on the web can help communicators stay abreast of misinformation.

“Fake news” has become a household term in recent years, so it’s no revelation that a lot of what goes viral is actually driven by bots, fake accounts, and coordination instead of organic interest. If you only tabulate likes and mentions for a topic, your picture of what’s “engaging” on the internet will be skewed by inauthentic data that shouldn’t be counted anyway.

You might also miss the larger point: Much of the apparent volume of conversation is driven by a proportionally tiny set of passionate users. Social media algorithms reward vocal, active, and influential accounts. As a result, a small number of social media handles — both authentic and inauthentic — tend to hold disproportionate sway over the conversation.

Yonder Narrative goes beyond just measuring engagement. Our goal is to form a nuanced picture of online conversation by understanding the behavior of factions, the influential online power groups driving viral stories and online narratives. Among the questions we try to answer:

  • Who originated the narrative and how widespread was its reach?
  • Did the narrative reach factions that don’t normally interact with each other, or did it stay close to its “ideological home”?
  • How likely is it to trend, and how might it impact brands, public sentiment, and related narratives?
  • What can the path of this narrative tell us about where online factions get their truth?

The nuances can be illuminating if, for example, you’re someone trying to understand why the company you work for is suddenly caught up in an unforeseen Twitter-gate, or carefully crafting the language for an upcoming press release.

While tracking Covid19 narratives to keep our clients ahead of misinformation and reputational threats, we caught up with Robert Matney, Yonder’s Managing Director of Web Authenticity, about how the sausage gets made when the news happens.

What’s step one for you when the news cycle goes haywire, as it did with Covid19? How do you and the team respond?

Just like we do with our clients, after our tool detected a rapidly-developing narrative, we tuned Yonder Narrative to focus very specifically on this topic and ran a machine-learning analysis to identify the various emerging terms people use when discussing Covid19. This lets us cast a wide enough net to be comprehensive, while remaining narrow enough to pick up what is most relevant.

There are infinite permutations of conversations and factions that we can track. To keep the results intelligible, how do you set up the parameters?

The good news is that once we cast that net targeting the most common terms, our tool automatically does a great job of surfacing the most important narratives related to a topic. As needed, we also factor the outputs of our platform (significant posts and narratives) back into tuning what we are looking for. That’s useful for really honing in on what matters.

How does Yonder Narrative automatically determine what’s authentic, and what sounds fishy?

Our tool is fantastic at identifying “bots and babies” (automated accounts and newly-created accounts, respectively) in a conversation; the tool measures and displays values for them. But the other part of conversation authenticity is post inequality, which is simply a measure of when a small subset of accounts drive an outsize proportion of the conversation. Of course, post inequality can result from sock-puppet accounts (a.k.a. “astroturfing”), but it can also be passionate, grassroots online activism that drives disproportionately high-volume account behavior. This spontaneous (or sometimes planned) coordination means that a small subset of accounts end up having outsize influence on the overall conversation. Our tool helps our customers understand when bots, babies, and post inequality impact the conversation. This way, customers have a sense of when a conversation is truly “one-person, one-voice”, versus a coordinated or amplified narrative.

Let’s talk about making predictions. How do we know when a narrative is likely to go viral as opposed to just remaining contained — albeit loudly — within a single faction?

We have metrics that pick up on unusual changes and volatility in engagement, and a time-series forecasting algorithm that looks at historical patterns to make predictions about future patterns. These give us an increasingly reliable sense of what’s coming around the corner.

Do the behavior patterns of factions we are seeing with Covid19 resemble other narratives we have seen in the past?

It won’t be a surprise to readers that online discussion is polarized along partisan lines. In the case of Covid19, Yonder Narrative showed that right- and left-leaning factions get their facts from distinctly different and unrelated sources. On this topic, the 35 most active factions cluster into roughly politically right-leaning and politically left-leaning interests and have virtually no mutual influence between clusters. The resulting conversational picture is even more polarized than usual:

Sample Map of Factions
This faction influence map shows how narratives about Covid19 flowed from faction to faction, January 21–February 12, 2020.
Each faction represented had a minimum of 2,000 posts about Covid19 during those dates. Engagements (replies, shares, retweets) are represented by arrows from the originating faction towards the engaging/amplifying faction. Arrows are only drawn when at least 5% of a faction’s posts are engaged with by that other faction. Size and color refer to a faction’s post count and conversational health/authenticity, respectively.

It’s also no surprise that factions most interested in the geographic site of the tragic health issue (factions that focus on China and associated issues) are heavily active in this topic. But seeing precisely where and how their influence is wielded on this topic adds context when considering strategic responses, such as public health outreach or communications campaigns.

When observing these patterns over time, the picture visible through Yonder Narrative can help shape decisions on product design and release planning, marketing and consumer research, and a host of other business questions.

See also:

More like this: