WILL THE DIGITAL SERVICES ACT PROMOTE SAFER AND HEALTHIER ALGORITHMIC RANKINGS?
summary
The DSA forces platforms to publish what parameters their recommenders use (Article 27: parameter transparency) and, for very large sites, to offer a non‑profiling alternative (Article 38: non‑profiling alternative); auditors and regulators currently lack data‑science capacity, researchers must wait for the delegated act on data access, and outside scraping is insufficient. Technical paths discussed include session‑only or explicit‑feedback recommenders, client‑side or encrypted computation like homomorphic encryption, and “bridge‑building” objectives instead of pure engagement metrics; defaults, A/B experimentation and label exposure are decisive for real impact.
AI-written summary · about
chapters (10)
- 0:00Recommender systems and DSA
- 4:15DSA provisions on recommenders
- 8:53Auditing DSA compliance challenges
- 13:52DSA versus GDPR in recommenders
- 21:15DSA impact on recommender ecosystem
- 23:23User control under DSA
- 36:12Reformer wishes for recommenders
- 50:01Behavioral inferences and user control
- 1:04:52Business models and hate speech under DSA
- 1:13:01Q&A on AI Act misuse
related talks
01h 15mSocial Media Recommender Systems Should Deliver Value, not “User Engagement"...
32mPlatform-supported Auditing Of Social Media Algorithms For Public Interest
01h 12mInvestigating algorithmic systems - Algorithm auditing.
01h 15mThe DSA as a tool in the fight against gender based violence
01h 06mEMPOWERING THE AI ACT: LIMITS AND OPPORTUNITIES
01h 12mEmerging Patterns in EU Digital Regulation
01h 18mApproaches to DSA Data Access
01h 19mAligning Data Privacy and AI Governance