An Ethics of Deconstruction responding to AI generated Bias

Abstract

The paper aims to respond to the ethical concern of biases generated by Artificial Intelligence systems. Even though biases enter an AI network via different channels, its presence in the algorithm can pose serious difficulties. AI systems have an algorithmic way of working where language is ‘formal’ and meaning is ‘fixed’. We employ deconstructive strategies of Jacques Derrida to understand the nature of this problem of AI bias through examination of algorithmic/programming language. Derridean philosophy looks at metaphysics as heavily depended on notions such as logocentrism where logos refers to the privileged part in a dichotomy. Logos is the centre of formal language which works as a system of signs. The main point of deconstruction is to apprise us of this privilege given to ‘presence’ of a concept or meaning over its ‘absence’. Derrida’s notion of Undecidables and Aporia, not only destabilizes rigid dichotomies like speech/writing but also give way to an opening of a concept in its ‘impossible -possibility’. In AI systems, programming/algorithmic language conforming to its algorithm offers definitive answers to problems. This takes us a step ahead in ‘formalization’ of language. Derrida, however offers a response via his ethical deconstruction where the process of ‘completion’ of any concept is deferred and its meaning ‘undecided’.

Ähnliche Arbeiten