Do Machines Have Prima Facie Duties?

Abstract

Which moral theory should be the basis of algorithmic artificial ethical agents? In a series of papers, Anderson and Anderson and Anderson (Proc AAAI, 2008[1]; AI Mag 28(4):15–26, 2007 [2]; Minds Mach 17(1)1–10, 2007 [3]) argue that the answer is W. D. Ross’s account of prima facie duties. The Andersons claim that Ross’s account best reflects the complexities of moral deliberation, incorporates the strengths of teleological and deontological approaches, and yet is superior to both of them insofar as it allows for “needed exceptions.” We argue that the Andersons are begging the question about “needed exceptions” and defend Satisficing Hedonistic Act Utilitarianism (SHAU). SHAU initially delivers results that are just as reflective, if not more reflective than, Ross’s account when it comes to the subtleties of moral decision-making. Furthermore, SHAU delivers the ‘right’ (that is, intuitively correct) judgments about well-established practical cases, reaching the same verdict as a prima facie duty-based ethic in the particular health-care case explored by the Andersons (a robot designed to know when to over-ride an elderly patient’s autonomy).