Promoting the need for truthful information to foster a democracy built on trust.
The Institute for Monitoring Emerging Inauthentic Narratives (l'Institut de Veille des Récits Emergents Inauthentiques, l'IVREI) is designed to monitor and analyse in real time narratives appearing on French-language social media.
Liberal democracies are under threat from information attacks by foreign powers, sometimes amplified by domestic actors. These attacks are carried out with the aim of polarising and fracturing French society, distorting the perceptions and opinions of the French public, and influencing the outcome of elections. Social media platforms are often a breeding ground for polarisation, as they expose users to radical minority views and to information that is false, misleading or hyper-partisan.
In response to the proliferation of these malicious operations and to prevent the artificial dissemination of narratives about current events, we have established the Institute for Monitoring Inauthentic Emerging Narratives. Our priority is to inform the media and decision-makers, in a fully transparent manner, of the risks of manipulation of public debate.
The IVREI is designed as an early warning system intended to:
In the short term, the objective is divided into two stages: first, to detect inauthentic narratives as they begin to go viral; second, to alert media professionals to prevent them from being disseminated to the general public. We produce monitoring reports (and alerts where necessary) for several national media outlets.
In the medium term, the data accumulated in this way will then be used to develop a predictive model of virality based on the stories detected. The aim is to gain a better understanding of disruptions to public debate on social media.
To achieve this, we combine artificial intelligence technologies — automated network analysis, computational text analysis, natural language processing and neural networks — with human expertise, to ensure real-time monitoring and analysis of social media.
This first stage involves identifying emerging themes within the French-speaking digital space. Topic modelling algorithms enable the automatic identification of topics gaining visibility, without any prior assumptions about their nature.
Within the identified themes, the team proceeds to identify specific, well-defined narratives. This detailed analysis makes it possible to distinguish between:
Stories identified as potentially inauthentic are subjected to a human assessment that examines several criteria:
