Asmaa
El Fraihi
I design experimental, data-driven methodologies to audit the algorithmic systems shaping our online world—measuring bias, opacity, and systemic risk in AI-driven platforms, with direct contributions to EU policy.
About
Building the tools to hold AI systems accountable.
I am a PhD candidate in Informatics, Data & AI at École Polytechnique and Inria (LIX laboratory), working under the supervision of Oana Goga. My research focuses on the algorithmic auditing of large-scale AI delivery and targeting systems.
Using black-box audit techniques—sock-puppet experiments, controlled-variation designs, and statistical inference—I investigate how platforms personalize content and ads, with a particular focus on transparency, fairness, and regulatory compliance under the EU Digital Services Act.
My empirical findings have directly informed policy submissions to the European Data Protection Board, translating research into regulatory impact.
Experience
Research & academic career
Feb 2023 – Jul 2026
Doctoral Researcher — AI & Platform Auditing
LIX / Inria · Palaiseau, France
- Designed and ran black-box audits of large-scale AI targeting systems using sock-puppet experiments and statistical inference to test for bias, opacity, and systemic risk.
- Built end-to-end measurement infrastructure quantifying cross-site user linking and tracking accuracy on Meta's advertising stack.
- Evaluated DSA-mandated transparency tools across major online platforms.
- Co-authored official policy feedback to the European Data Protection Board (Guidelines 3/2025).
- Co-supervised junior researchers on platform-auditing projects.
Feb 2026 – Apr 2026
Visiting Research Fellow
Max Planck Institute for Software Systems (MPI-SWS) · Kaiserslautern, Germany
- Designed and deployed a GDPR-compliant data donation platform for collecting user conversation logs across five major conversational AI assistants.
- Conducted comparative empirical analysis across Gemini, Claude, DeepSeek, Grok, and Mistral to characterize cross-platform differences.
Oct 2021 – Jul 2022
Research Intern — Master's Thesis
LIG (Grenoble Computer Science Laboratory) · Grenoble, France
- Built NLP and text-classification pipelines (transformers, lexical-resource models) to characterize persuasion signals in online advertisements at scale.
- Designed large-scale controlled online experiments on Meta's AI-based ad-delivery personalization system.
Oct 2017 – Oct 2022
Computer Science Engineering Diploma & Research Master's
ESI Alger (École nationale Supérieure d'Informatique) · Algeria
- Software Engineering major; Research Master's in Computer Science.
Research
Selected publications
Understanding Conversational AI Through Everyday Use
An Experimental Framework for Auditing Ad Delivery Responsiveness to Psychological Traits ↗
To appear in the International AAAI Conference on Web and Social Media (ICWSM)
Is Contextual Advertising Safe? Analyzing Systemic Risks with Ads on YouTube ↗
The ACM Web Conference (WWW) 🏅 Inria-VIGINUM Young Researcher Award
A Year Under the DSA: Ad Transparency's Uneven Landscape ↗
Privacy Enhancing Technologies (PoPETs)
How Persuasive Are LLMs in the Wild? Assessing Personalized Ads in Real-World Delivery ↗
International AAAI Conference on Web and Social Media (ICWSM)
Client-side and Server-side Tracking on Meta: Effectiveness and Accuracy ↗
Privacy Enhancing Technologies (PoPETs) 🏆 Best Student Paper Award
Policy engagement
Feedback to the EDPB's Guidelines 3/2025 on the DSA–GDPR interplay ↗
L. Zard, O. Goga, A. El Fraihi, N. Bielova · Submitted to the EDPB public consultation, 2025.
Provided empirical evidence and methodological input to EU discussions on digital-platform accountability.