The man behind Trump’s Twitter account
Turning unwieldy social media data into actionable intelligence.
Have you ever heard of Dan Scavino? Well, if it makes you feel any better, neither had I until I dived deep into Trump’s now extinct twitter feed. But trust me, you should know who he is!
A recent Politico piece dubbed Scavino the ‘ultimate insider’ in the Trump administration, Steve Bannon has referred to him as “essentially the comms department of the White House”, and he’s one of the only people who still remains in Trump’s inner circle after the tumultuous last four years of Republican presidency.
Rumour has it that Trump regularly bent Scavino’s ear for strategic advice on issues such as whether the US should withdraw troops from Syria, what Trump should tweet this morning, and even hardline immigration policies.
To top all of this off, a 16-year-old Scavino first met Trump when he was a golf caddie. Since then, his career has segued from running Trump’s golf club in Westchester County, NY, to being Trump’s senior adviser for digital strategy.
From a political analyst’s point of view, Scavino is a gift that just keeps on giving. So how did I find out about this mysterious guru and social media chief of Trump’s 2016 campaign? Well, since you ask, it all began with a data science course I took at General Assembly. I wanted to study up on my python and data skills to formalise a lot of the professional skills we use in our social media monitoring work here at PGI.
As a final project, I chose to analyse Trump’s tweets – which have thankfully been archived online by a wise and prescient developer. I was interested to see if the data would show the driving forces of Trump’s divisive speech. Is Trump’s rhetoric more divisive and hateful if he’s been firing off virtual backhanders all morning with Nancy Pelosi? Or does he really only get fired up about illegal immigrants when Brietbart slides into his retweets?
To do this, I got a lot of extra data from the Twitter API since I wanted to see how citations from domains like Breitbart affected Trump’s language. I then built a hate speech classifier using an independent hate dataset and filtered the dataset for tweets from 2016 only, i.e. after Trump became president.
Now, this is where things get statistical. I was interested in which features have an outsized impact on Trump’s divisiveness. I had a numerical score for the divisiveness of each tweet, generated by that hate speech classifier I mentioned above, so I conducted a p value test with a simple linear regression algorithm to determine which other twitter accounts or web domains had an outsized impact on the divisiveness of Trump’s tweets.
In plain English: I tested whether any twitter account or website made Trump really extremely unusually divisive or the opposite – unusually quiet and demure. That’s what you can see in this graph below:
A lot of this is line with what you’d expect. For instance, at the bottom left of the graph, you can see that Trump frequently disagrees with Bloomberg – denoted here by ‘bloom’ since they often shorten urls to bloom.bg on Twitter. But if you dig a little deeper…
… you might just find Dan Scavino! He’s right in the middle and he influences Trump’s divisiveness in both directions. The point here is that Scavino has an outsized impact on Trump’s Twitter feed, which is what the p value test was supposed to determine. When I looked a little deeper into Scavino, I found that he had been Trump’s Twitter guru since the 2016 presidential campaign and even has his own following of tech-savvy Trump supporters, particularly on Reddit.
It may not seem like a major thing, but it very much underscores how a contextually nuanced approached to social media monitoring can be fruitful in the extreme. The data doesn’t lie and you can pry tremendous insights out of it such as this discovery of Dan Scavino – but you have to then also go and read the Politico article to understand who Scavino is and why it’s important!
This is just the sort of work we do for clients on a day to day basis. We often pull fascinating insights out of otherwise large and unwieldy social media data to help our clients understand who the real drivers of hate are, mis- and disinformation online and how you can act accordingly.
Subscribe to our Digital Threat Digest, insights from the PGI team into disinformation, misinformation, and online harms.
PGI’s Digital Investigations Team work with both public and private sector entities to help them understand how social media can affect their business. From high level assessments of the risks of disinformation to electoral integrity in central Africa to deep dives into specific state-sponsored activity in eastern Europe we have applied our in-house capability globally.
Contact us to talk about your requirements.
Online influence campaigns are becoming increasingly common as political parties and state actors around the world seek to manipulate public opinion.
To most people, online influence operations involve competing ideologies battling it out in the public sphere.
Last week, Russian President Vladimir Putin complained that former Fox News anchor Tucker Carlson had been too soft; saying Carlson avoided “sharp questions” during their interview on 06 February.