Webinar Recap: Open Source Intelligence Tools and Trump
During this presidential election, we have seen a dramatic increase in the use of social media by public officials. From state senates to the White House, politicians are embracing the use of posting on these platforms to express their viewpoints, accomplishments, criticisms, and personal opinions. The use of open source intelligence tools helps determine words that come from political staffers and which come straight from the politicians themselves.
Arguably, the most discussed example of this trend comes from the Twitter account of presidential nominee, Donald Trump. Trump tweets frequently about a variety of different topics that are considered controversial. For better or worse, his Twitter activity helps make him the most talked about person in the world today.
In this webinar, we discussed how BrightPlanet harvested 3,200 tweets from Trump’s Twitter account for personality profiling. The tweets were passed along to John Kreindler and Sean Farrell of Receptiviti, a psychology-based data collection company, in order to uncover which tweets are created from a Trump staffer and which tweets are authentically Trump’s own words. Through analysis of the language, Receptiviti was able to dictate the emotion, tone, and decision-making processes behind each published tweet. Here’s what they found:
Trump’s Tweets Come from Two Devices
One of the major discoveries made was that Trump tweets came from an Android device and an iPhone. This is significant, because there are major differences between the tweets that come from each device. With their NLP software uncovering people’s psychology through their tone and speaking patterns, Receptiviti notes that the Android tweets are full of exaggerated statements and hyperbole in comparison with the iPhone tweets, which tend to be more neutral.
The Words Make the Person
Our words make us who we are. David Robinson, a data scientist at Stack Overflow, notes that the publishing of the Android tweets were much earlier in the day. He also notes, the Android tweets have a tendency to contain fewer links and hashtags, simpler words, and were angrier in tone. Based on comparing tweets to past transcripts, we’ve found that our model can classify Android/iPhone personalities within 94% accuracy. Overall, the Trump press conference transcripts were very similar to the Android tweets with 74% votes in similarity.
We Could Determine the iPhone Staffer
After identifying Trump as the likely author of the Android tweets, the question that remained was who was the author of the iPhone tweets? We examined nine of Trump’s top staffers that are active on Twitter. As active users, they all have their own distinct personalities and individual characteristics when it comes to their speech. The current RF model has the ability to categorize staffers over 89% of the time. Comparatively, the voice of each staffers had their own speech, tone, and display of critical-thinking. We found a staffer who was a match, with 53% similarity to the iPhone tweets. The profiles of Gavin Smith, the South Carolina Field Director for Trump’s campaign were similar to the iPhone Trump tweets.
Open source intelligence, as demonstrated by BrightPlanet and Receptiviti, can open up a whole new world of information. Download the free webinar to learn more.
//
