Who are the "agents" in game theory, do they apply to AI in computer games?

SpaceDweller October 27, 2022 at 09:39 1700 views 2 comments

I have a question related to game theory for which regardless of your expertise I'm willing to hear your rational though and personal opinions regardless of your position, so please don't hesitate to comment.

In search for definition of "Game theory" I come upon several different definitions, following is the definition from Stanford encyclopedia:

Game theory is the study of the ways in which interacting choices of economic agents produce outcomes with respect to the preferences (or utilities) of those agents

As per OP title I'm interested in those "economic agents", that is to whom do they apply?

If you scroll down to "2.1 Utility" it says this:

An economic agent is, by definition, an entity with preferences. Game theorists, like economists and philosophers studying rational decision-making, describe these by means of an abstract concept called utility.

https://plato.stanford.edu/entries/game-theory/

Another definition is ex. from wikipedia which defines game theory like this:

Game theory is the study of mathematical models of strategic interactions among rational agents

On wikipedia they are called "Rational agents" instead of "economic agents" and are further defined on wikipedia as:

A rational agent or rational being is a person or entity that always aims to perform optimal actions based on given premises and information.

https://en.wikipedia.org/wiki/Game_theory

For simplicity let's call them just "agents" (or players)...

According to these definitions agents are rational beings or entities, with preferences and ability to make rational decisions right?

My question is simple:
Can we consider AI in computer games such as strategy game, where a human player competes against one or multiple AI's to be an agent?

I'm asking this because there are some arguments against AI qualifiying as an agent. ex:

- AI in computer games isn't perfect due to complexity of games and ingenuity of human player, and many times it doesn't make rational decisions.
- AIs do not get tired or require food and toilet breaks.
- AI is just an algorithm and not human.
- Games are emotionally stressful for humans but not for AI.
- AI can have significant advantages over human players
and similar...

Beside human and AI in games, to whom else could these agents apply?

Comments (2)

Metaphysician Undercover October 27, 2022 at 11:02 #751922 0 likes

I believe the "agent" is anything you want it to be in game theory, so long as it is acting. This is why game theory has a very wide range of applicability. You could probably even adapt game theory to be applied to quantum physics, where fundamental particles are the agents involved in decision making.

SpaceDweller October 27, 2022 at 11:58 #751929 0 likes

Quoting Metaphysician Undercover

I believe the "agent" is anything you want it to be in game theory, so long as it is acting. This is why game theory has a very wide range of applicability.

Thank you for reply.

I must admit I didn't read trough the whole article in the link I posted, but I did read a bit more now in trying to understand what you said and I see where does this lead to...

There are basically 2 camps of opinions as to who agents are by interpreting "utility",
one camp which defines utility by way of RPT (Revealed Preference Theory), and other camp which doesn't:

RPT camp:

Economists and others who interpret game theory in terms of RPT should not think of game theory as in any way an empirical account of the motivations of some flesh-and-blood actors (such as actual people).

Non RPT camp:

Some other theorists understand the point of game theory differently. They view game theory as providing an explanatory account of actual human strategic reasoning processes.

Further the continuation of the article seems to lean toward RPT camp:

An economically rational player is one who can:
1. assess outcomes, in the sense of rank-ordering them with respect to their contributions to her welfare
2. calculate paths to outcomes, in the sense of recognizing which sequences of actions are probabilistically associated with which outcomes
3. select actions from sets of alternatives (which well describe as choosing actions) that yield her most-preferred outcomes, given the actions of the other players.

Which excludes human player as the only one qualifying as "rational agent":

An entity is usefully modeled as an economically rational agent to the extent that it has alternatives, and chooses from amongst these in a way that is motivated, at least more often than not, by what seems best for its purposes.

Thus according to RPT camp definition, it qualifies the AI in computer games as rational agent.

Therefore I guess those arguments against AI in my OP can be dismissed if non RPT stance is also dismissed as definition of game theory, but that's up to debate because I'm sure non RPT camp has it's own arguments to defend their stance that only human beings qualify.