The Right to Robot Identification : Using Cryptographic Hashes to Manage AI Black-Boxes

Interacting with an AI system, a “robot”, as in financial robo-advice, can pose significant challenges from the end-user and the algorithm developer’s perspective. Robots’ decisions can be hard to explain and justify, let alone assure that they are free of biases. Furthermore, the most advanced algorithmic decision-making machine learning models can continuously evolve, learning from their mistakes and absorbing new training data, something that makes their accountability even more elusive. If someone believes they have been wronged in algorithmic decision-making, they have quite limited resources using current legal liability schemes. In this work, we propose and analyse novel “crypto-legal” tools, that enable a reliable way for an individual to precisely identify the robot behind any algorithmic decision they receive, as well as obtain assurances about their certification, if the machine learning algorithms are certified. The method we propose provides a way towards the accountability of the AI system: on the one hand, it enhances the ability of an individual to have tangible and admissible evidence in court to support their claim, and on the other it provides a way for service providers, such as Fintechs, to demonstrate compliance at the point of service and in case of a lawsuit