Navigating the social dilemma of autonomous systems: normative and applied arguments

Abstract

Autonomous systems (ASs) become ubiquitous in society. For one specific ethical challenge, normative discussions are scarce: the social dilemma of autonomous systems (SDAS). This dilemma was assessed in empirical studies on autonomous vehicles (AVs). Many people generally agree to a utilitarian programming of ASs, but do not want to buy a machine that might sacrifice them deterministically. One possible way to mitigate the SDAS would be for ASs to randomize between options of action. This would bridge between a socially accepted program and the urge of potential AS users for some sense of self-protection. However, the normativity of randomization has not yet been evaluated for dilemmas between self-preservation and self-sacrifice for the “greater good” of saving several other lives. This paper closes this gap. It provides an overview of the most prominent normative and applied arguments for all three options of action in the dilemmas of interest: self-sacrifice, self-preservation, and randomization. As a prerequisite for inclusion in societal discussions on AS programming, it is ascertained that a normative argument can be elicited for each potential course of action in abstract thought experiments. The paper then progresses to discuss factors that may shift the normative claim between self-sacrifice, self-preservation, and randomization in the case of AV programming. The factors identified in this comparison are generalized into guiding dimensions for moral considerations along which all three options of action should be evaluated when programming ASs for dilemmas involving their users.