From BADs to BEDs. Algorithmic Disclosure Regulation. Theoretical aspects for empirical application

Amidst the growing skepticism surrounding transparency measures, this paper supports that disclosure regulation can play a role in online transactions, if it is designed through algorithms. This requires solving several theoretical legal challenges and bridging technical solutions, making this article contribute to the Law&Tech scholarship. The contribution of this article is threefold. First, it provides a review of an emerging scholarship: the Law&Technology (Law&Tech), which is distinct from the Technology or Innovation Law, the Legal Tech and Empirical Legal Studies. Law&Tech is a new research field that uses tools from Natural Language Processing (NLP) and Machine Learning (ML) to investigate on legal topics. We classify it around four research fields: (i) analysis, (ii) interpretation (iii) application and enforcement, (iv) enhancement. (Part One) Second, the article draws on the fourth (iv) future-oriented sub-category of the Law&Tech literature to demonstrate that research on how algorithms could contribute to enhance the rule- making of disclosure regulation is still limited. It therefore advances the current knowledge on how that would be possible. Third, it presents a new (two-phase) model to draft effective disclosure regulation. In both phases, algorithmic technologies are used, making this contribution fit the Law&Tech scholarship. In phase one, we suggest using Natural Language Processing and Machine Learning technologies to first combine Rules and Industry disclosures (the de iure and de facto disclosures), and then grade them to select those that ‘fail’ the least: the BADs, or Best Available Disclosures. (Part Two) In phase two, we propose using regulatory sandbox to test the BADs and get to the BEDs (or Best Ever Disclosures). Technically speaking, the BEDs would: be adaptive to the addressees’ informational needs (i.e. feed-in through behavioral data); targeted at diversified groups of addressees; cheap to implement. From a legal point of view, testing the messages in the sandbox before implementation serves to grant due process (transparency and participation) of algorithmic rule making; but also to increase BEDs’ proportionality (least possible burdensome solution for recipients) (Part Three). Keywords: Law, disclosure duties, disclosure regulation, natural language processing, machine learning, data, competition, regulatory sandbox