Pilot project lighthouse: A proposed GDPR compliant methodology for analysing special categories of personal data

Abstract

The General Data Protection Regulation (GDPR) is designed, in part, to prevent discrimination in algorithmic decision making. However, the GDPR's requirements, as well as EU member states' implementing laws, often make testing for bias using special categories of data, such as race, either impractical or impossible. This paper argues that the pilot project lighthouse methodology is a GDPR compliant method for bias-testing special categories of data in algorithms. This paper finds that the pilot project lighthouse methodology is permissible in the majority of EU member states and argues that to the extent pilot project lighthouse methodology would be prohibited by either the GDPR or an individual member state's implementing laws, the same are contrary to the letter and intent of the GDPR.