What is an agent, and how can its interaction with an environment be described? Start with the general perception–action framework, then explore a specific model of decision-making under partial observability.
- Agent — the perception–action framework, agent functions and programs, rationality, and task environments.
- Action — a choice available to a decision-maker.
- Observation — evidence received about a state.
- Belief state — a probability distribution over possible hidden states.
- Policy — a rule for selecting actions from available information.
Read the scope of each entry: Action, Observation, Belief state, and Policy use a specific POMDP formulation. For a different meaning of belief, see Theory of mind. For how an LLM system supplies information to a decision step, continue to Memory and information.
For limits on deliberation, design, and the organization of decisions, continue to Design and organizational decision-making.