A Metrological Approach to Ethical and Legal Issues in Artificial Intelligence

摘要

Artificial Intelligence has developed in an impressive way during the recent years, and is now being applied to almost every field of human activities, slowly replacing human beings in operations whose level of required skills has significantly increased. Collaborative robots, or cobots, are a reality in industrial production, as well as virtual reality and robots driven by human motions from remote sites allow operators to control operations in dangerous areas. AI algorithms perform data searches and present the results in a very efficient way, so that they are helping decision makers in critical fields, such as medicine and justice. This poses new and somehow unforeseen ethical and legal problems that must be covered to avoid generating wrong or even illegal results. Some of these wrong results might be generated by the use of input data that might not be sufficiently accurate, especially when they are collected from the field, or whose limited accuracy is not properly considered when processing them. This paper aims at considering a possible, metrologically-sound approach to ethical and legal issues met in AI.