The Digital Services Act’s Article 27 will require online platforms to inform their users about the main parameters used in their recommender systems and of the options to tweak those parameters. On the other hand, Article 38 will require the largest online platforms to provide an alternate recommender system not based on profiling as defined in the GDPR. What impact will this have on polarisation, hate speech, data protection risks, or other behavioural consequences that stem from personalised recommendations optimised for engagement? How will this affect transparency and user control?
The DSA already arrives at a time of increased discussion of chronological news feeds and research into alternative recommendation models, for instance through the opportunities of open-source and local/community governed models or through techniques like “bridging”. It also comes at a moment of heightened scrutiny of social media platform recommender systems and efforts make them more transparent and accountable.
• Will the DSA truly create space for alternatives, and how will we assess these alternatives?
• How can a satisfactory user profile be constructed and maintained while respecting the regulatory requirements and balancing potentially conflicting interests?
This discussion will weigh the technical possibilities for widespread uptake and debate the potential individual and societal benefits as well as the risks and trade-offs.