Online platforms: users under influence
In 2018, Cambridge Analytica made headlines. The London-based company developed a computer program capable of predicting and potentially influencing the choice of American voters at the time of the 2016 presidential elections. By taking a personality test on Facebook, 270,000 users of the platform had their personal data and interaction history siphoned off. Unknowingly, they opened access to information about their friends and enabled the psychological profiling of 87 million users. This data was then used in Donald Trump's 2016 election campaign to target messages according to profiles and increase the chances of swinging the election in his favor.
Influence of targeting technologies
In 2018, Cambridge Analytica made headlines. The London-based company developed a computer program capable of predicting and potentially influencing the choice of American voters at the time of the 2016 presidential elections. By taking a personality test on Facebook, 270,000 users of the platform had their personal data and interaction history siphoned off. Unknowingly, they opened access to information about their friends and enabled the psychological profiling of 87 million users. This data was then used in Donald Trump's 2016 election campaign to target messages according to profiles and increase the chances of swinging the election in his favor.
Influence of targeting technologies
"Targeting technologies - powered by algorithms, themselves driven by artificial intelligence - were initially used to sell products. The Cambridge Analytica scandal has shown that they are now being applied to information, paving the way for the manipulation of users of online platforms (Google, Facebook, Tik Tok...)", points out Oana Goga, research fellow at École Polytechnique’s Computer science laboratory (LIX*).
The European Union is also taking the issue very seriously. Fearing the militarization of technology or the instrumentalization of voters, it took the researcher's work into account when drafting the Digital Services Act of 2022 and, through it, demanding greater transparency from platforms.
In the same year, the European Research Commission awarded the scientist a five-year, €1.5M ERC Starting Grants award for her Momentous project (Measuring and Mitigating Risks of AI-driven Information Targeting) dedicated to the study of risks associated with information targeting. "Initially, we are seeking to determine the extent to which a person's cognitive biases (rapid, short-cut reasoning, leading to hasty decisions) can be used to influence their opinions and preferences. To do this, we rely on randomized controlled trials and technology that combines computer science and behavioral economics," explains Oana Goga. The Momentous team will then turn the spotlight on targeting algorithms to determine what they are capable of learning on their own, about the vulnerabilities and cognitive biases of platform users. Finally, the last part of the project will study how these targeting technologies impact the quality of information received and assimilated by users.
Auditing online platforms
To carry out the Momentous project, Oana Goga is drawing on the expertise she has developed in studying the dangers of online platforms for societies and democracies. In the wake of the Cambridge Analytica affair and shortly before the 2018 Brazilian presidential elections, the researcher had conducted a study among local Facebook users. She had provided them with a tool for collecting the ads they had received while an algorithm distinguished between political and non-political content. "This method of analysis and measurement makes it possible to audit the targeting techniques implemented by platforms and the risks associated with them. It uses data donation, a particularly innovative concept in IT research". Thanks to this process, Oana Goga was able to demonstrate that the Facebook Political Ad Library was not displaying all the political ads produced at the time, despite announcements of transparency, and presented a risk of influencing voters.
The work of the Momentous ERC is part of this dynamic, and will ultimately provide a solid basis for technologies guaranteeing healthy and sustainable targeting of information delivered on social network
*LIX : a joint research unit CNRS, École Polytechnique, Institut Polytechnique de Paris, 91120 Palaiseau, France
Further informations :