Was Trump Conducting A/B Testing With His Crazed Tweets?
On October 5th, just a few days after testing positive for COVID-19, the USA president Donald Trump’s Twitter was ablaze with a series of what appeared to be his run-of-the-mill deranged tweets — those we have grown used to (exhausted by?) over the last four if not eleven (he joined in 2009) years.
Each tweet appeared almost like a Python dictionary where the key was an issue of supposed import to conservatives, and the value was the exclamation “VOTE.”
But wait, there’s more:
Now while one could look at what we have above and truly believe this came from the mind of Donald Trump, a man then-high on a cocktail of Dexamethasone, Remdesivir, Adderall, and Lord-knows-what-else — there could be a more intentional mind at play. This series could have been Trump’s team A/B testing topics for the president to hit in the final month leading up to the election.
I imagine Stephen Miller’s intern, tasked to run Trump’s Twitter account while the president remains high out of his mind, writing something like this:
Although, if we’re honest, this is Stephen Miller’s hypothetical intern we are talking about, it probably started as something like this:
Ok, but what is A/B testing and why might Trump and/or his team engage in it?
A/B testing, also known as split testing, is a form of hypothesis testing. A researcher shows a group of people two variants of the same variable, and the researcher examines the responses. It is carried out across the sciences and arts and is especially common in advertising. In the case of Trump and his tweets, it would be more like A/B/C/D/E/F/G/etc. testing.
Generally, when A/B testing, a researcher has a null hypothesis and an alternative hypothesis. A researcher then examines whether the group’s response to either variant of a particular variable determines whether or not they should reject or fail to reject their null hypothesis.
In the case of Trump’s team, they very well could have created this long string of tweets and examined the responses from Twitter in many ways. The possibilities ranged from a cursory examination and recording of likes vs. retweets vs. comments (to suss out if any particular tweet may have been “ratioed”) to a natural language processing of each tweet’s responses to delve explore how positive/negative folks’ responses were.
I have not yet delved into whether or not Trump decided to emphasize those topics from the tweets in this list that were particularly popular by some metric; however, it would be interesting to analyze in the future.
I’m not sure whether I would ascribe a whole lot of strategy to what we have seen out of the final months of Trump’s campaign, so perhaps, this was Trump ignoring any semblance of a plan as per usual.