Free will and intelligent machines: Why case-based reasoning systems do what they do

摘要

This paper presents an analysis of the application of the philosophical concept of free will to intelligent machines. It is argued that we need to transfer the concept to a more empirical approach in philosophy in order to be able to answer questions about the design of intelligent systems that are required to operate in areas of society where the concept of free will is necessary. Therefore, free will is viewed as a social or relational concept. The relations between intelligent machines and their environment are defined in terms of interaction points on the knowledge level and the dialogue level. Based on a case study into case-based reasoning and the CREEK system developed at NTNU in Trondheim, these relations are analysed from both an actor-network and a phenomenological perspective.According to the actor-network approach, we can inscribe properties into artefacts. This leads to a framework for modelling free will in intelligent machines, founded on the tension between what a machine ‘can’ do and what it ‘will’ do, or on the connection between hermeneutic and existential views in phenomenology. Phenomenology also provides the possibility for evaluation of the mediation machines can perform in the relation between humans and reality, which is necessary if we are introducing machines in decision-making. This view is linked to the perspective of free will issues. Some evaluative remarks are given based on the analysis of case-based reasoning and CREEK.