I apologize if this thread doesn't belong here but it seemed like a good place to start. I'm working on the final project for my math's degree on the topic of belief revision, mostly based on the approach described on Sven Ove Hanson's article called "Revision of Belief Sets and Belief Bases". One of the concepts that is constantly mentioned throughout the text but is never explicitly defined is "agent". I more or less understand what it stands for, which appears to be "anything that has the ability to execute an algorithm". However, I would like to give a more precise definition and I just don't know where to look. It doesn't have to (and actually shouldn't) be too technical. I only need a concise definition that gives some context to the word. Ideally I would also like to provide a reference for it, so I would be grateful if you could point me to some article or book where I can find this information.
Philosophically speaking, 'agency' is a euphemism for the concept of free will, a euphemism meant to sidestep all of the knotty problems that arise whenever one talks about free will. With that in mind, an 'agent' is anything that senses its environment and makes (implicitly non-deterministic) alterations in its state. That parenthetical is necessary — we need it to distinguish simple mechanical sensors from beings that are capable of conscious decision-making — but it is left undiscussed in many cases because authors would like to avoid metaphysical entanglements.
Put in more simple terms, we have discussions bounded by these constraints:
- No one wants to say that a mercury thermostat or photoelectric sensor is the same 'kind' or 'class' of object as a human (or presumptive AI) consciousness, even though the former technically make 'decisions'
- Many people want to avoid invoking the concept of a soul, or of any transcendent, metaphysical 'essence' that constitutes the capacity for free will, thought, or reason; sometimes because they distrust those concepts and sometimes because those concepts are superfluous and distract from their main point
'Agency' is a quick way of pointing at the capacity to make non-deterministic decisions without getting bogged in the tar-baby of the metaphysics of consciousness.
I think the book "Artificial Intelligence, A Modern Approach" by Stuart Russel and Peter Norvig gives a good definition. In my 1995 edition, right at the beginning of chapter 2 "An agent is everything that can be viewed as perceiving it's environment through sensors and acting upon that environment through effectors".
A useful distinction can be drawn between an agent and a free agent.
A plain agent is any entity which receives information, process it and acts on the outcome. Software agents are, or were, common in hierarchical software applications, where an agent would typically run on a client host and interact with the main service on a remote system.
A free agent embodies the philosophical notion of free will, being able to make decisions and think up its own actions, independently from any inputs received.
The fact that the term is not defined in the text you ask about makes it difficult to know how (or even whether) the author was handling this distinction.