Algorithmic Disclosure Regulation

This project brings together an interdisciplinary group of legal experts and data scientists with the aim of exploring how technology could improve disclosure regulation. The research group seeks to develop and test examples of data-driven, targeted and partially automatized regulatory processes to make rule-making more efficient and the resulting norms more effective.

We investigate the potential of Machine Learning tools in overcoming the failure of countless disclosures (e.g. privacy notices, terms and agreements, risk information) in truly informing consumers about the issues they want to know about.

The envisioned outcome of this research is to better understand how technology could help to increase the efficiency of this part of law.

Design and test an algorithmic process of developing algorithmic disclosure regulation. Algorithms are being used very successfully by firms to tailor their online offers and advertisement strategies to the individual’s preferences. Thanks to a continuous data flow, the user is always presented with what she is eventually interested in.

The Algorithmic Disclosure Regulation project seeks to apply the same concept to clusters of consumers who share similar capacities and informational preferences regarding disclosure statements.

Our work complements existing approaches in this field by offering a uniquely holistic approach, covering both the rulemaking and the implementation phase.

Furthermore, we investigate how co-regulatory instruments, which bring together regulators, industry representatives, and consumers, could help to make this process more efficient and future-proof. As cross-cutting issues, questions of due process and democratic legitimacy, which are of utmost relevance when relying on algorithmic tools, are addressed.

Our research questions:

• Drafting Stage

1. Why do disclosures fail to inform and empower its addressees? How can we measure these failures?

2. Can we automize the process of linking provisions of general disclosure regulations (e.g. the GDPR) to disclosures drafted by firms and presented to consumers (e.g. a privacy policy)?

• Interaction phase

1. How can we assess which disclosures perform well in practice, i.e. whether they are read and understood by the addressee?

2. How can we make disclosures more effective and proportionate? Can we automatically adjust disclosures to the specific informational needs of certain groups of addressees?

• Implementation phase

1. Could we design an automatic, data-driven implementation process for differentiated disclosures?

§How can we lower the cost of producing and updating effective disclosures?

• Cross-cutting questions

1. What way can co-regulatory processes be integrated in an algorithmic disclosure scheme?

2. How can we guarantee due process, accountability, and democratic legitimacy?

Our research group seeks to explore the potential of algorithmic disclosure with a view to several sectors: potential fields of application include

• consumer protection law,

• data protection law,

• health law,

• financial law, and

• competition law.

Thanks to our interdisciplinary group, we will use both traditional methods of doctrinal legal research as well as data science methods, especially from the field of Machine Learning and NLP as well as behavioral sciences.

We seek to yield the following outputs:

• Develop a comprehensive, theoretical algorithmic disclosure framework

• Implement this framework on a small-scale, prototype test run

• Share and discuss both the framework and test results

• Implement it on large scale

Our overarching goal is to contribute to making disclosure more effective and thus empower disclosure addressees while alleviating businesses from complicated disclosure drafting processes.