A lesson from AI: Ethics is not an imitation game

Abstract

How can we teach moral standards of behavior to a machine? One of the most common warnings against AI is the need to avoid bias in ethically-loaded decision-making, even if the population on which the learning is based is itself biased. This is especially relevant when we consider that equity (and protection of minorities) is an ethical notion that by itself goes beyond the (most probably) biased opinions of people: equity must be pursued and ensured by social structures, regardless of whether people agree or not. We know (believe?) that bias, or being biased, is a bad thing, regardless of what the majority says. In other words, good and evil is not what the majority says, it is beyond majorities and mathematical formulae. Ethics cannot be based on a majority opinion about right and wrong or on a rigid code of conduct. We need to overcome the generalized skepticism in our society about the rationality of ethics and values. The good news is AI is forcing us to think ethics in a new way. The attempt to formalize ethics in a set of rules misses the point that a person is not only an instance of a case, but a unique and unrepeatable being. Ethics should prevent us from the error of converting equity into mathematical equality, achieved through the extraction of characteristics and the computation of a value formula. Equity is not mathematical equality, not even a weighted equality that considers different factors.

The title of the 2014 movie The Imitation Game tells us the life of Alan Turing, especially his outstanding participation in the decipherment of the German messages encrypted with the Enigma machine in the Bletchley Park Complex [1] , [2] . The expression “the imitation game” is from Turing himself: these are the first words of his 1950 article, Computing Machinery and Intelligence [3] . It is also the name of a game played by the Victorian aristocracy, which consisted in a blind exchange of handwritten messages to try to guess whether the interlocutor was a woman or a man.