Média Animation and the Conseil supérieur de l’éducation aux médias (CSEM) have collaborated to produce a series of webinars exploring the question: “Is AI redefining the challenges posed by disinformation?”. In this fourth and final instalment of the series, Jérémy Grosman discusses the issues surrounding recommendation algorithms and their link to the challenges of disinformation and polarisation.
Following his presentation, a round-table discussion between media literacy professionals was organised using a collaborative approach, with a view to exploring the topic in greater depth and putting it into practice.
Understanding algorithms and their effects on disinformation
Before delving further into the issues surrounding these technologies, Jérémy provides a detailed explanation of what algorithms are. In a constant stream of online content, algorithms enable information to be prioritised and categorised, allowing users to engage with content that matches their interests.
He also puts their impact on disinformation into perspective. Indeed, according to a study carried out on the social media platform Facebook and explained by Jérémy, the use of a chronological filter would make false information more visible than recommendation algorithms, which can reduce exposure to disinformation content by more than 60 per cent.
A study of Trump’s 2016 campaign also examines the link between the mainstream media and social media. It shows that false information may first appear on social media before being picked up by the mainstream media. It is only at this second stage that spikes in search queries appear on social media.
Online disinformation can therefore exist in a latent form and be amplified when it is given a public platform. This highlights an intrinsic feature of the internet: anyone can become a source of information.
Conceptual limits of disinformation and polarisation
Jérémy goes on to define the concepts of disinformation and polarisation, which are often taken for granted, whilst highlighting their limitations.
In his words, “disinformation is a concept that essentially characterises the process of communicating information, based on the correction of the information received: the information is deemed to be incorrect or incomplete.” In his view, the concept of disinformation does not adequately account for the contextual process of creating and disseminating such disinformation.
Polarisation, for its part, is defined by the distance separating individuals or groups of people based on their stances – particularly political ones – or their expectations. This concept does, however, have certain limitations.
Intuitively, one might assume that extreme or radical views are necessarily problematic. However, according to the researcher, their existence can also be a sign of a healthy democracy, in which different viewpoints can be expressed and debated.
The problem arises when those defending a view are no longer able to respond to the objections raised against them. This can happen when the different parties in a debate are not on the same level of discussion. For example, whilst one person is trying to discuss a specific issue, the other may shift the discussion towards a question of values. The arguments put forward are therefore no longer on the same level and become difficult to compare directly. This is what Jérémy illustrates through the distinction between first order and second-order problems.
Introduction to the tool
The teaching sheet offers a simplified explanation of algorithms and puts algorithmic bubbles into perspective. It then provides a set of questions to help teachers spark a discussion with their pupils about their algorithmic environment.
To find out more about the content of the sheet, click here.