An AI agent is an autonomous computational entity that interacts with other entities or an environment and adapts its behavior. It may act alone or cooperate with other agents to achieve complex goals without centralized control. section 9.1 section 12.1
Common agent settings differ in how agents interact, learn, and pursue goals:
- Swarm agents are relatively simple, spatially distributed entities whose local interactions produce decentralized, self-organized collective behavior. section 9.1
- Reinforcement-learning agents map situations to actions, learn through trial and error, and use rewards or penalties to maximize cumulative reward. section 12.1
- MDP agents account for states, transitions, and consequences across multiple time steps; multi-armed-bandit agents choose among actions without explicit states or long-term planning. section 12.1
Foraging and collective decision-making
Foraging illustrates agent behavior at two scales: individual organisms respond to environmental signals, while groups coordinate labor, protect members, or search collectively without a central controller. Optimization methods model these patterns through algorithms inspired by bacteria, birds, fish, ants, bees, and other organisms. section 9.1
