Different approaches to the moral status of AI: a comparative analysis of paradigmatic trends in Science and Technology Studies

Abstract The exponential progress of AI systems today compels scientists and philosophers to redefine their conceptual frameworks to better understand the nature of these new technologies and their moral status. Among the various theories that are used to respond to the challenges posed by intelligent systems are instrumentalism, Socio-technical Systems Theory (STST) and Mediation Theory (MT), all widely adopted in the field of Science and Technology Studies (STS). This paper intends to present the main features of these theories and provide a comparative analysis of them in order to assess their contribution to the process of understanding the moral status of artificial intelligence. Our investigation intends to show how (1) instrumentalism is inadequate to account for the moral status of AI, (2) STST, while helping to highlight the link between AI, society and morality, lends itself to the criticism of anthropocentrism, (3) MT in its Latourian version has the merit of highlighting the active character of technological artefacts and thus of artificial intelligence in the moral sphere. However, the principle of symmetry it proposes poses the problem of the de-accountability of the human agent. (4) MT in its postphenomenological form seems to partially resolve the problem of moral responsibility, but the opacity of the terminology it employs exposes it to various criticisms. In light of these results, we intend to show how an understanding of the moral status of intelligent systems cannot be based on the diametrically opposed positions that consider technologies either morally neutral or else moral agents similar to humans, whereas particularly useful elements can be found in STST and in postphenomenological MT.