We tried to give a sketch of the answer using artificial intelligence to analyze millions of tweets.
When talking about the Trumpian phenomenon we can distinguish three parts: the context, the leader and his crowd. In the context front, folks have explained how globalization and technology have pushed waves of unschooled blue-collar workers out of the labor market; the global rejection of mainstream politics is also evoked. Several people have described Donald Trump as a textbook populist, others paint him as an authoritarian. Although some work has been done at analyzing Trump supporters, it has been restricted to demographics and small focus groups. We decided to focus on Trump’s crowd taking a more massive approach.
What did we do?
We gathered four million tweets belonging to more than two thousand hard-core Trump supporters. We’re convinced that we can better grasp a community’s picture of the world by understanding their dialect. Languages are the framework of our thinking. If we get to understand how words are associated inside a person’s mind, we’ll understand how her ideas came to be.
To analyze all tweets efficiently, we trained an artificial intelligence to create a new representation of words. The AI encoded the meaning of words as dense high-dimensional vectors inside a semantic space. Distances between those vectors encoded the semantic distance between their associated words (e.g. the vector representation of the word morons was near idiots but far away from funny). Crucially, the map that transformed words into a vectors was estimated from the data. So, we could analyze what concepts were associated to which ideas within the community from which we gathered the data.
How to make sense of it?
The artificial intelligence we trained gave us a powerful quantitative model of the language used by Trump supporters. Now, the challenge we faced was to define what questions we wanted to answer. To formulate these questions we used a mix of linguistics and political science.
Patrick Charaudeau, a renamed French linguistics researcher, posits that the purpose of politicians actions are both to keep their followers support and to swell their ranks.
The classic schema used by politicians is dubbed the Triadic Scenario and it has three steps:
- The politician presents the current state as a complete disaster, and makes the audience feel as if they were the victims.
- Then the politician exhibits a common enemy: the origin of all problems and suffering.
- Finally the politician indicates that there is a way to solve all problems, and that he is the only one being able to bring it to life.
We found Trump’s speech and strategy astonishingly similar to the theory posed by Professor Charaudeau. So, we decided to analyze the Trumpian phenomenon using his framework.
What messages echo in Trump’s crowd?
We started our analysis by understanding what were the most echoed ideas inside Trump’s crowd. For that, we used our tweet corpus and thousands of tweets sent from Trump’s official account. We counted the meme repetition frequency and their virality.
An impressive portion of the most viral terms were the ones in which Donald Trump attacked his competitors. The only popular meme related to public policy was the one related to illegal immigration.
The data let us see very clearly that Donald Trump’s is doing a good job defining common enemies within his community. It also let us see who their enemies are: politicians and immigrants, especially Mexicans.
Why is Trump being so successful at positioning Mexicans as enemies?
We used our AI to retrieve the 100 closest neighbours to the word Mexicans. Then, we mapped the high-dimensional vectors that encoded the meaning of the given words into a 2D space. The algorithm used to map the high dimensional words representation into a low dimensional space was designed to visualize high dimensional data preserving the distances between the different data points. Thanks to this method, it is possible to visualise in two dimensions, the location of the different terms associated to Mexicans.
The figure shows the terms in the neighborhood of the word “Mexicans” in the semantic space. We can distinguish four themes: Immigration, Crime, Big government and Jobs. The theme that contains the more words is Crime.
Let us analyse the semantic map above in the light of Professor Charaudeau theory. The agreement between the classic schema of political persuasion and a portion of the semantic neighborhood of Trump Supporters is just astonishing.
- A common enemy: Mexicans and immigrants in a broader sense, also government.
- A disastrous state of affairs: Crime and unemployment.
The semantic space surrounding the term Mexicans depicts a terrifying world. The semantic space of Donald Trump followers is ruled by fear. Trump supporters seem to be afraid of losing their jobs and they also seem to be obsessed with the idea of becoming the next victims of crime. The semantic space has two important regions, just in between the two oceans of fear: The Big government and the Immigration zones. These zones are related to the enemy, the origin of all evil.
