What Do 19.95 Million Covid-19 Tweets Teach Us About The Future Of Social Media?

Apr 14 2020

Today, when I type “Covid-19” into the trusty google search bar, I’m offered over 2B results. As in billion. When you look at the Google Trends map, all this content has been queried, generated and shared in just a few short months, if not weeks (surprisingly, especially among the curious residents of South Dakota!). So it seems worth better understanding where this information is emanating from and how trustworthy it is - especially on social media, where this kind of information flies even faster.

The New York Times recently reported a sudden surge in misleading posts on Twitter and Facebook related to the coronavirus and Covid-19, which led data science expert, and my colleague at AI services company KUNGFU.AI, Dr. Steve Kramer, to conduct an analysis on Twitter using 19.95M tweets that mention Coronavirus and/or COVID-19. He discovered some important and surprising things regarding how disinformation works its way into our feeds. You can find a more detailed view of his work and findings in this blog post, but for a quick view, we also spent some time on video talking through the implications of his analysis, especially for business leaders.

In brief, Steve’s two key findings:  

1. Judged by the posts which get the most engagement (likes/comments/shares), the winning combo are those generated by cyborgs: humans who use special software to amplify and automate their social posts - meaning that for now at least, we are less compelled by messaging created by fully automated bots. Just as chess champion Garry Kasporov and others have learned, human/AI collaborations win over over algorithms alone.  

2. Two of the three twitter domains with the greatest number of URLs linked to by the automated accounts spreading that this virus is “just the flu”, are influential Russian news websites - which given this pandemic is a global issue is notable (even if we can’t discern if these accounts are controlled by state actors).

My additional takeaways:

  • The use of human/bot cyborgs can be harnessed for good (as well as malintent), as many of the “fake news” reports came from what we judge as credible news outlets. It means this approach will gain acceptance by brands eager to efficiently amp up their messaging.
  • This is especially important given how easy it would be to hijack and discredit authentic brand messaging only generated by humans. The AI enabled social content tug-of-war will no doubt grow.
  • Which is why we all have a responsibility to verify a message before we reflexively hit “like” or “share”.
  • And we should better understand the policies of the platforms we choose to use. We should vigilantly advocate greater care be taken, especially as advertisers supporting these platforms.    

Given this analysis, Facebook-owned instant messaging service, Whatsapp announced last week that any message that has been forwarded five or more times will now face a new limit that will prevent a user from forwarding it to more than one chat (contact) at a time. This builds on the move they made in 2018 to lower the numbers from 20 to five, allowing it to reduce the volume of message forwards globally by 25%. And as misinformation and racist jokes were spreading throughout India, Twitter released two blog posts on how they are actively working to dampen this by more aggressively using AL and machine learning to screen out harmful information... and by broadening the word “harm”.  

Meanwhile, I was intrigued to learn it is possible to utilize free data like this! As we said above, Steve was able to analyze more than 19.95M tweets that mention Coronavirus and/or COVID-19 using the free Twitter public API that offers access to a small (~0.5%) sample of the full Twitter firehose. He has now posted part of this data set on data.world (a modern catalogue for data analysis) and is inviting data science collaborators to continue monitoring this issue, not least because as we learned in 2016, the implications have a significant impact on the upcoming US election.

I encourage business leaders in all organizations to consider the powerful impact — both good and threatening — of AI/ML amplified social media messaging. Using AI/ML is a critical part of deciphering our way through the sheer volume of misinformation proliferating social channels, especially when we are actively seeking new information on how to keep ourselves, families and communities safe. Credible sources must be able to share news fast. And it’s clear... trust is becoming a brand’s most important asset.