Much of my research involves creating intelligent systems that ``inhabit'' a synthetic environment, generating human-like behavior and possibly learning to improve that behavior over time. I have always felt comfortable describing such a system as an ``agent''.
In the mid-1990s, a new label arrived on the scene of research in artificial intelligence: ``intelligent agent''. It seems to me that this label is very descriptive of the type of systems I build. However, this label has come to describe a different class of systems.
The potential confusion arises from competing definitions of the word ``agent''. My definition was something like ``an autonomous entity that inhabits a realistic environment''. However, the ``intelligent agent'' label uses a definition like ``an entity that helps one accomplish ones goals'', such as a real estate agent, or a theatrical agent. Some of the differences between these definitions may be subtle, but in terms of current research into artificial intelligence, they describe very different classes of systems.
Thus, I have adopted the preference of describing my research systems as ``intelligent actors'' rather than ``intelligent agents''. This new label accomplishes a number of goals: