David Robinson is a data scientist at Stack Overflow, a Q&A website for software developers.
A data scientist just analyzed over a thousand tweets from Donald Trump's Twitter account. His findings may help point out when it's actually Trump tweeting, as opposed to a staffer. (Jhaan Elker/The Washington Post)

Recently, political and digital analysts began to suspect that Donald Trump’s tweets are authored by two different people: Someone on his campaign staff is tweeting from an iPhone, and the billionaire himself is tweeting from his Samsung Galaxy. Some posts are markedly more hyperbolic and aggressive (“Like the worthless @NYDailyNews, looks like @politico will be going out of business. Bad reporting- no money, no cred!”) than the campaign boilerplate (“Thank you Windham, New Hampshire! #TrumpPence16 #MAGA”). In short, they sound like Trump. When Trump wishes the Olympic team good luck, it comes from an iPhone. When he’s insulting a rival, it usually comes from an Android.

My recent work has concerned text-mining and sentiment analysis, so I wanted to quantify the difference in Trump’s tweets. The data clearly shows that the Android and iPhone tweets are from different people, posting at different times of the day, and using hashtags, links and retweets in distinct ways. What’s more, the Android tweets are angrier and more negative, while the iPhone tweets tend to be benign announcements and pictures. This survey looks at 628 iPhone tweets and 762 Android tweets.

One consideration is what time of day the tweets occur, which we’d expect to be a signature of their author.

Another major difference involves the sharing of links or pictures in tweets. This is a standard practice among social-media professionals of the kind Trump would employ, but it’s uncommon on his own phone.

Tweets from the iPhone were 38 times as likely to contain a picture or a link, often as part of announcements about rallies or other events.

By filtering the text through some analysis software, we can also see what kind of language comes from each phone:

Emotionally charged words such as “badly,” “crazy,” “weak” and “dumb” are overwhelmingly more common on Android. I also found that most hashtags come from the iPhone; Trump himself almost never uses them.

Then I ran this set of tweets through the National Research Council Canada’s Word-Emotion Association Lexicon, which links words with 10 sentiments: positive, negative, anger, anticipation, disgust, fear, joy, sadness, surprise and trust. Trump’s Android account uses 40 to 80 percent more words related to disgust, sadness, fear, anger and other negative sentiments than the iPhone account does. (The usage of words linked to positive emotions wasn’t different to a statistically significant extent.)

A longer explanation the methodology can be found here.