Periodic Reporting for period 1 - MOMENTOUS (Measuring and Mitigating Risks of AI-driven Information Targeting)
Reporting period: 2022-10-01 to 2025-03-31
We are witnessing a massive shift in the way people consume information. In the past, people had an active role in selecting the news they read. More recently, the information started to appear on people’s social media feeds as a byproduct of one’s social relations. At present, we see a new shift brought by the emergence of online advertising platforms where third parties can pay ad platforms to show specific information to particular groups of people through paid targeted ads. These targeting technologies are powered by AI- driven algorithms. Using these technologies to promote information, rather than to promote products as they were initially designed for, opens the way for interested groups to use user’s personal data to manipulate them. European institutions recognize the risks, and many fear a weaponization of the technology to engineer polarization or manipulate voters.
The goal of this project is to study the risks with AI-driven information targeting at three levels: (1) human-level–in which conditions targeted information can influence an individual’s beliefs; (2) algorithmic- level–in which conditions AI-driven targeting algorithms can exploit people’s vulnerabilities; and (3) platform- level–are targeting technologies leading to biases in the quality of information different groups of people receive and assimilate. Then, we will use this understanding to propose protection mechanisms for platforms, regulators, and users.
This proposal’s key asset is the novel measurement methodology I propose that will allow for a rigorous and realistic evaluation of risks by enabling randomized controlled trials in social media. The measurement methodology builds on advances in multiple disciplines and takes advantage of our recent breakthrough in designing independent auditing systems for social media advertising. Successful execution will provide a solid foundation for sustainable targeting technologies that ensure healthy information targeting.
The goal of this project is to study the risks with AI-driven information targeting at three levels: (1) human-level–in which conditions targeted information can influence an individual’s beliefs; (2) algorithmic- level–in which conditions AI-driven targeting algorithms can exploit people’s vulnerabilities; and (3) platform- level–are targeting technologies leading to biases in the quality of information different groups of people receive and assimilate. Then, we will use this understanding to propose protection mechanisms for platforms, regulators, and users.
This proposal’s key asset is the novel measurement methodology I propose that will allow for a rigorous and realistic evaluation of risks by enabling randomized controlled trials in social media. The measurement methodology builds on advances in multiple disciplines and takes advantage of our recent breakthrough in designing independent auditing systems for social media advertising. Successful execution will provide a solid foundation for sustainable targeting technologies that ensure healthy information targeting.
Contribution 1: Uncovering exposure and consumption of (mis)information in social media
Social media has become an important platform for news publishers to promote their work and engage with their readers. A long-standing question in the field has been how much misinformation users see in their timelines and how they engage with this content. Previous works could not provide answers due to the lack of access to data on the content users see in their timelines. In [4], we designed CheckMyNews, a software that can collect the posts related to news sources users are exposed to and how users interact with them on social media. The paper is the first to quantify the amount of misinformation in the feeds of 500 U.S. participants. It provides a nuanced perspective on the underlying mechanisms governing news appearance. It is also the first paper that challenges current beliefs, showing that users are willing to engage with posts from opposing views, provided their interactions are not visible to their friends.
We see an emergence of inauthentic websites claiming to be news media that have attempted to influence voters during elections. In [pre1], we propose a method to detect self-proclaimed news providers automatically. We discovered five times more news providers in the U.S. than previously known.
Contribution 2: Understanding marketing to children online
In [3] this paper, we investigated whether it is technically possible to target children with ads online and analyzed the legal requirements regarding marketing to children online. The paper shows that although there are no direct ways to reach children, YouTube allows advertisers to instruct the platform to only place their ads on precise videos or channels–so advertisers can market to children by simply placing their ads on videos "made for kids". We created several ad campaigns targeting children-focused videos, and they all reached users. On the legal side, our analysis shows that both COPPA (U.S.) and DSA (E.U.) forbid advertising to children based on user profiling but do not forbid contextual advertising. This is problematic as ads targeted at a specific video are hard to verify by parents or civil society; this was not the case on ads placed on children-focused programs on TV. We also showed that YouTube allows advertisers to combine placement and interest-based advertising. This might be a violation of the COPPA and the 2019 YouTube settlement.
Contribution 3: Privacy risks with new tracking technologies
One core objective of the proposal is to understand the risks of information targeting through online advertising.
To increase users' privacy, all major browser vendors have deprecated, or plan to deprecate, third-party cookies to reduce tracking. Despite these efforts, advertising companies continuously innovate to overcome restrictions. Recently, advertising platforms, like Meta, have been promoting server-side tracking solutions to bypass traditional browser-based tracking restrictions. Our paper [5] explores how server-side tracking technologies can link website visitors with their user accounts on Meta products. Our findings show that Meta’s server-side technology can match between 34% and 51% of website visitors to user profiles on Meta products using basic information like the visitor’s IP address, user agent, and location data.
[pre1] Automated Discovery of Self-Proclaimed News Providers on Facebook. S. Chouaki, M. Nguyen, L. Edelson, T. Lauinger, D. McCoy and O. Goga 2024
[6] Weaponizing the Wall: The Role of Sponsored News in Spreading Propaganda on Facebook D. D. Singh, G. Chauhan, MM. Nguyen, O. Goga, A. Chakraborty. International Conference on Advances in Social Networks Analysis and Mining (ASONAM), Sept. 2024 (B)
[5] Client-side and Server-side Tracking on Meta: Effectiveness and Accuracy A. Elfraihi*, N. Amieur*, W. Rudametkin, O. Goga Privacy Enhancing Technologies Symposium (PETS), July 2024 (A)
[4] What News Do People Get on Social Media? Analyzing Exposure and Consumption of News through Data Donations S. Chouaki, A. Chakraborty, O. Goga, S. Zannettou ACM The Web Conference (WWW), April 2024 (A*)
[3] Marketing to Children Through Online Targeted Advertising: Targeting Mechanisms and Legal Aspects T. Medjkoune, O. Goga, J. Senechal ACM Conference on Computer and Communications Security (CCS), November 2023 (A*)
[2] Understanding the Privacy Risks of Popular Search Engine Advertising Systems Salim Chouaki, O. Goga, H. Haddadi, P. Snyder ACM Internet Measurement Conference (IMC), November 2023 (A)
[1] On Detecting Policy-Related Political Ads: An Exploratory Analysis of Meta Ads during the 2022 French Election V. Sosnovik, R. Kessi, M. Coavoux, O. Goga ACM The Web Conference (WWW), April 2023 (A*)
Social media has become an important platform for news publishers to promote their work and engage with their readers. A long-standing question in the field has been how much misinformation users see in their timelines and how they engage with this content. Previous works could not provide answers due to the lack of access to data on the content users see in their timelines. In [4], we designed CheckMyNews, a software that can collect the posts related to news sources users are exposed to and how users interact with them on social media. The paper is the first to quantify the amount of misinformation in the feeds of 500 U.S. participants. It provides a nuanced perspective on the underlying mechanisms governing news appearance. It is also the first paper that challenges current beliefs, showing that users are willing to engage with posts from opposing views, provided their interactions are not visible to their friends.
We see an emergence of inauthentic websites claiming to be news media that have attempted to influence voters during elections. In [pre1], we propose a method to detect self-proclaimed news providers automatically. We discovered five times more news providers in the U.S. than previously known.
Contribution 2: Understanding marketing to children online
In [3] this paper, we investigated whether it is technically possible to target children with ads online and analyzed the legal requirements regarding marketing to children online. The paper shows that although there are no direct ways to reach children, YouTube allows advertisers to instruct the platform to only place their ads on precise videos or channels–so advertisers can market to children by simply placing their ads on videos "made for kids". We created several ad campaigns targeting children-focused videos, and they all reached users. On the legal side, our analysis shows that both COPPA (U.S.) and DSA (E.U.) forbid advertising to children based on user profiling but do not forbid contextual advertising. This is problematic as ads targeted at a specific video are hard to verify by parents or civil society; this was not the case on ads placed on children-focused programs on TV. We also showed that YouTube allows advertisers to combine placement and interest-based advertising. This might be a violation of the COPPA and the 2019 YouTube settlement.
Contribution 3: Privacy risks with new tracking technologies
One core objective of the proposal is to understand the risks of information targeting through online advertising.
To increase users' privacy, all major browser vendors have deprecated, or plan to deprecate, third-party cookies to reduce tracking. Despite these efforts, advertising companies continuously innovate to overcome restrictions. Recently, advertising platforms, like Meta, have been promoting server-side tracking solutions to bypass traditional browser-based tracking restrictions. Our paper [5] explores how server-side tracking technologies can link website visitors with their user accounts on Meta products. Our findings show that Meta’s server-side technology can match between 34% and 51% of website visitors to user profiles on Meta products using basic information like the visitor’s IP address, user agent, and location data.
[pre1] Automated Discovery of Self-Proclaimed News Providers on Facebook. S. Chouaki, M. Nguyen, L. Edelson, T. Lauinger, D. McCoy and O. Goga 2024
[6] Weaponizing the Wall: The Role of Sponsored News in Spreading Propaganda on Facebook D. D. Singh, G. Chauhan, MM. Nguyen, O. Goga, A. Chakraborty. International Conference on Advances in Social Networks Analysis and Mining (ASONAM), Sept. 2024 (B)
[5] Client-side and Server-side Tracking on Meta: Effectiveness and Accuracy A. Elfraihi*, N. Amieur*, W. Rudametkin, O. Goga Privacy Enhancing Technologies Symposium (PETS), July 2024 (A)
[4] What News Do People Get on Social Media? Analyzing Exposure and Consumption of News through Data Donations S. Chouaki, A. Chakraborty, O. Goga, S. Zannettou ACM The Web Conference (WWW), April 2024 (A*)
[3] Marketing to Children Through Online Targeted Advertising: Targeting Mechanisms and Legal Aspects T. Medjkoune, O. Goga, J. Senechal ACM Conference on Computer and Communications Security (CCS), November 2023 (A*)
[2] Understanding the Privacy Risks of Popular Search Engine Advertising Systems Salim Chouaki, O. Goga, H. Haddadi, P. Snyder ACM Internet Measurement Conference (IMC), November 2023 (A)
[1] On Detecting Policy-Related Political Ads: An Exploratory Analysis of Meta Ads during the 2022 French Election V. Sosnovik, R. Kessi, M. Coavoux, O. Goga ACM The Web Conference (WWW), April 2023 (A*)
There are many research groups and organizations that are interested in developing our measurement methodology that allows to measure user's exposure, consumption and engagement with information on online platforms (CheckMyNews, [4]).
The paper [3] unveils how children can be targeted with ads online, and it shows major gaps in the current legislation. We are trying to lobby for more protective regulations on marketing to children.
The work in [5] shows how online platforms adapt to circumvent tracking protections from browsers. It is important to understand its effectiveness and accuracy and raise awareness on the limits of current tracking protections with respect to new technological developments.
The paper [3] unveils how children can be targeted with ads online, and it shows major gaps in the current legislation. We are trying to lobby for more protective regulations on marketing to children.
The work in [5] shows how online platforms adapt to circumvent tracking protections from browsers. It is important to understand its effectiveness and accuracy and raise awareness on the limits of current tracking protections with respect to new technological developments.