Constrained Incrementalist Moral Decision Making for a Biologically Inspired Cognitive Architecture

The field of machine ethics has emerged in response to the development of autonomous artificial agents with the ability to interact with human beings, or to produce changes in the environment which can affect humans (Allen, Varner, & Zinser, 2000). Such agents, whether physical (robots) or virtual (software agents) need a mechanism for moral decision making in order to ensure that their actions are always beneficial, that they ‘do the morally right thing’. There has been considerable debate on what doing the right thing means, and on how moral decision making should be implemented (Anderson & Anderson, 2011; Lin, Abney, & Bekey, 2011; Wallach & Allen, 2008) in order to create so-called Artificial Moral Agents (AMAs) (Allen et al., 2000). Apart from the problem that no consensus on ethics exists, it has also proven to be exceedingly difficult to computationally implement the often vague and under-constrained