For this edition, we've asked Oana Goga, Chargée de Recherches at CNRS, three questions to shed light on some of her recent research and upcoming plans.

Can you tell us a bit about your project MOMENTOUS: Measuring and Mitigating Risks of AI-driven Information Targeting? 

The idea of the MOMENTOUS project came when I realised that targeting mechanisms used in online advertising have started to be used to target information instead of products. 

What I am trying to understand is whether we can pinpoint and quantify in which conditions "targeted information" can influence a user's opinion. For example, to what extent can one exploit the cognitive biases of users to convince them to vote for a particular policy? Many of us have heard about Cambridge Analytica and the fact they tried to influence voters before the 2016 U.S. presidential elections using messages that appealed to users' personalities. But, to this day, there is no scientifically sound evidence that it was effective. This is where my project comes in; we proposed a measurement methodology based on randomized controlled trials and user data donations to assess the effectiveness of information targeting. The second aspect we look at is whether AI-driven algorithms that decide who sees what ad can learn by themselves the vulnerabilities of users and serve them ads with information presented in a way they are more likely to accept. 

The DSA includes a provision to give access to platform and search engine data for researchers fulfilling certain conditions. How do you hope your research could benefit from this initiative? 

I have been looking at risks with online platforms for nearly 15 years now, and I can't express enough how huge this provision is for us. Together with my students, I have spent an enormous amount of time developing ways to collect & grab data from online platforms to study risks. 

One question we urgently need answers to is exposure and consumption of misinformation. We know misinformation is present, but we do not know to which extent it reaches users and through which mechanism.  Given a user's timeline (be it on Facebook, TikTok, or Instagram), I would like to know how many posts are news-related, how many of them come from sources known to be factual, and how many come from sources known to have repeatedly spread misinformation. I would also like to know whether the post appeared in the users' timeline because it was recommended by the platform, because a friend shared it, because an advertiser paid the platform to show that information, or whether the user just followed the source. We need to understand the dissemination mechanisms to build safer technology. 

Where can we find out more about MOMENTOUS, the full range of your work and stay up to date as you progress?

Please check my personal webpage: https://www.lix.polytechnique.fr/~goga/. I am putting all the research papers I have published there. 

I have worked on risks related to disinformation, child protection, political advertising, advertising transparency, privacy violations, and security risks.