Third edition of Artificial Intelligence: foundations of computational agents, Cambridge University Press, 2023 is now available (including the full text).

1.5.6 Number of Agents

An agent reasoning about what it should do in an environment where it is the only agent is difficult enough. However, reasoning about what to do when there are other agents who are also reasoning is much more difficult. An agent in a multiagent setting should reason strategically about other agents; the other agents may act to trick or manipulate the agent or may be available to cooperate with the agent. With multiple agents, is often optimal to act randomly because other agents can exploit deterministic strategies. Even when the agents are cooperating and have a common goal, the problem of coordination and communication makes multiagent reasoning more challenging. However, many domains contain multiple agents and ignoring other agents' strategic reasoning is not always the best way for an agent to reason.

Taking the point of view of a single agent, the number of agents dimension considers whether the agent does

  • single agent reasoning, where the agent assumes that any other agents are just part of the environment. This is a reasonable assumption if there are no other agents or if the other agents are not going to change what they do based on the agent's action.
  • multiple agent reasoning, where the agent takes the reasoning of other agents into account. This happens when there are other intelligent agents whose goals or preferences depend, in part, on what the agent does or if the agent must communicate with other agents.

Reasoning in the presence of other agents is much more difficult if the agents can act simultaneously or if the environment is only partially observable. Multiagent systems are considered in Chapter 10.