Samenvatting

Constructing artificial moral agents serves at least two purposes: one, better understanding of moral reasoning, and, two, increasing our trust and confidence in creating autonomous agents acting on our behalf. When constructing agents we must identify what kind of agent we are trying to implement acknowledging still severe restrictions in both our understanding of moral reasoning and technological capabilities. In constructing agents, as part of the SophoLab project, we adopt a set of requirements and design principles. As building blocks the belief-desire-intention model is used along with the deontic-epistemic-action logic framework as modeling and specification languages. Implementation in executable software is done using the JACK agent language. Attempts at constructing agents and actual experiments provide some first encouraging results but also point at the large hurdles still to be taken.