On Admissible Behaviours for Goal-Oriented Decision-Making of Value-Aware Agents

Abstract

The emerging field of value awareness engineering claims that software agents and systems should be value-aware, i.e. they should be able to explicitly reason about the value-alignment of their actions. Values are often modelled as preferences over states or actions which are then extended to plans. In this paper, we examine the effect of different groundings of values depending on context and claim that they can be used to prune the space of courses of actions that are aligned with them. We put forward several notions of such value-admissible behaviours and illustrate them in the domain of water distribution.