“Openness won’t hurt: Enforcing qualified machine learning transparency through responsive regulation“.
machine learning, transparency, black box, accountability, responsive regulation.
Machine-learning (ML) models have been increasingly applied to make decisions that affect key aspects of people’s lives. However, users and regulators are barely aware of how these models work, as only scarce information is disclosed by developers and operators on this matter. ML transparency emerges thus as a recurrent demand made by stakeholders for users to gain control over how much their lives should rely on judgements carried out by machines, for regulators to render those responsible for them accountable for incurred damages and for scholars to understand algorithms' impacts in society. This dissertation thus traces a comparative analysis on how the Brazilian and European data protection legal frameworks address ML transparency and assesses the adequateness of the responsive regulation theory’s participatory strategies and incentives framework for promoting more intelligible systems.