Workshop on Algorithmic Governance in the Platform Economy
The 2nd Expert Group for the Online Platform Economy held an online workshop on September 20th, 2022, as a part of Workstream 5: Algorithmic Governance in the Platform Economy. The main questions explored pertained to the patterns or processes that lead or may lead to discriminatory outcomes for groups of individuals on large social media platforms, focusing on algorithmic design, content moderation, and risk management both from academic and industry perspectives.
To this end, after an introduction by the Head of Unit at the European Commission, Prabhat Agarwal, the context and aims of the workshop were presented by Dr Celine Castets-Renard, member of the Expert Group and co-leader of the workstream alongside Dr Teresa Rodriquez de las Heras Ballell.
This was followed by the first part of the workshop, which focused on insights from academia. First, Dr Joris van Hoboken (Vrije Universiteit Brussels/’Fundamental Rights and Digital Transformation’) shed light on the challenge of connecting risk- and rights-based approaches in algorithmic governance, the intersection of the Digital Services Act (DSA) with algorithmic governance, and potential future scenarios. Thereafter, Dr Carlos Castillo (Universitat Pompeu Fabra) delved into the complexities of how biases may occur in data, the risks arising from using AI for content moderation, and the tendency for online hate to come in waves. The third presentation from the academic perspective was delivered by Dr Thomas Davidson (Rutgers University) who explored the problems associated with detecting online hate speech, how to measure and mitigate bias, and policy implications for the European Commission in light of this information.
Industry practices were presented during the second part of the workshop. Namely, Gemma Galdon-Clavell (Eticas Consulting) spoke of the different business models and incentive systems of industry players, accountability and measures to close the enforcement gap in the ecosystem of digital products.
In the final part of the workshop, the research carried out by the Observatory, as well as points for future consideration were showcased. Visionary Analytics (contractor) proposed a definition of algorithmic discrimination and a typology of content moderation as a part of the literature review, coupled with three case studies of discriminatory outcomes that have occurred on large social media platforms in recent years. Dr Cristiana Markou (member of the Expert Group, European University Cyprus) then touched upon protected speech law and immunity as well as the DSA provisions concerning the over-removal of content and transparency.
Throughout the workshop, the members of the Expert Group and the invited speakers engaged in Q&A sessions after each presentation and discussed in more detail the nature and risks of algorithmic discrimination, mitigation measures, and the way forward for European policymaking. Specifically, the main outcomes of the workshop concerned discussions on how to best reconcile risk-based and rights-based approaches to content moderation, as well as reflections about the importance of building an effective risk assessment mechanism. In this respect, several suggestions for a better digital products ecosystem were presented, including the creation of algorithm registries, leaflets and scores, and the introduction of audits and regulations of access to information.