Foundations of Rational Agency

The word agent is widely used in computer science these days. It can mean anything from a few lines of code that gets executed automatically when a certain condition is satisfied (e.g., a daemon that gets scheduled by the Operating System every day at midnight), to a more sophisticated program that can reason about its own behaviour and achieve long-term goals (e.g., an autonomous intelligent vehicle on Mars) [13]. In addition to the simplicity or sophistication of these software entities, agents are also characterised into different types based on their functionality. Interface agents, mobile agents, believable agents, reactive agents, learning agents, and emergent agents are just a few of these types. The level of autonomy (from being completely autonomous to being intelligent assistants to humans) and the level of their interaction with physical entities (from being totally in the software environment — softbots — to being intimately connected with their physical environment — robots) are also used to characterise agents.