The detection of conspiracy theories within reach of an algorithm? –

Swiss and British researchers have shown that conspiratorial productions (5G, vaccines, Illuminati, etc.) stand out from the point of view of automatic language processing by their lack of consistency.

What distinguishes a conspiratorial production from a classic informative text? Researchers from the universities of Neuchâtel and Warwick (United Kingdom) have just shown that they have a different structure which could lead to their automatic identification, in an article published on October 26, 2022 in the review Science Advances. We asked psychology researcher Adrian Bangerter (Unine), who led this work, to comment on it.

Adrian Bangerter is a professor at the Institute of Work and Organizational Psychology at the University of Neuchâtel. He is particularly interested in popular beliefs and their dissemination. | Unine, courtesy – Can you put this work into context?

Adrian Bangerter- The first author of the study, Alessandro Miani, a doctoral student in psychology at our Institute of Work and Organizational Psychology, studies the linguistic characteristics of conspiratorial texts. There are a lot of works on the fake news and their automatic identification by algorithms, but not much about conspiracy theories. We started with develop a large corpus of texts collected on the web, approximately 96,000 web pages and 80 million words.

How was this conspiratorial corpus formed?

We started with “germ” words that could give rise to conspiracy theories, such as “Princess Diana” or “coronavirus” for example, used in investigations on the subject. From these words, we identified conspiratorial texts, based on a selection of sites labeled as highly conspiratorial by Media Bias / Fact Check. For example, there is the Info Wars site of the American conspirator Alex Jones, recently sentenced to pay nearly $1 billion in damages to the victims of the mass shooting at the Sandy Hook school in Connecticut, which he claimed never happened. This is a classic example of a conspiratorial site.

Your hypothesis is that conspiratorial texts tend to mix several heterogeneous subjects?

There is a certain “promiscuity” in the theories, you can get a good informal idea if you know the world of conspiracy. People who believe in a conspiracy theory also tend to believe in others of the same type, and link them together.

For example, we talk about the covid vaccine and we make the link with the 5G antennas, themselves linked to the empire of Bill Gates, which is an Illuminati… We have several keywords connected in a way, say, creative. This is well established by opinion polls, when people are asked to indicate their level of belief in different conspiracy theories.

Reserved for subscribers

This article is for subscribers only.

Already subscribed?
To log in

We would love to thank the writer of this write-up for this awesome material

The detection of conspiracy theories within reach of an algorithm? –

Visit our social media accounts along with other related pages