Aarhus University Seal / Aarhus Universitets segl

What principles for moral machines?

Arto Laitinen, University of Tampere, Finland

This paper examines four kinds of principles for machine morality. (1) exceptionless principles covering all cases and features (e.g. Kantianism, consequentialism); 2) a plurality of midlevel prima facie -principles each concerning one act-in-a-situation type (Ross; Beauchamp&Childress; applied to robotics by Anderson&Anderson), as well as their typical enabling and disabling conditions; 3) priority principles concerning the midlevel principles (e.g. F. M. Kamm; the ordering of Asimov's Laws); 4), overall judgements in situations when everything relevant is taken into account and the situation is closed by a “that’s it”-consideration; stressed by particularists (Dancy, cf. Guarini). The paper argues that principles of type (2) are epistemologically most robust and they are possibly useful even in the absence of overall judgements of type (4).