Swarm Intelligence

Swarm intelligence is the collective behavioral capacity that emerges when large numbers of simple agents interact locally, producing coordinated global behavior without centralized control or any individual agent's comprehension of the collective outcome. The canonical biological examples are ant colonies, termite mounds, and murmuration of starlings; the canonical machine implementations are ant colony optimization, particle swarm optimization, and swarm robotics. The key empirical finding: the computational power of a swarm routinely exceeds the sum of its individual agents' capacities. A single ant implements roughly a dozen behavioral rules; an ant colony solves optimization problems — shortest-path routing, load distribution, task allocation — that would require sophisticated planning from a centralized reasoner. Swarm intelligence systems implement group-level selection explicitly: fitness is evaluated at the collective level, not the individual. This makes them a natural laboratory for testing whether Multi-Level Selection dynamics generate adaptations inaccessible to individual-level optimization. The field's foundational challenge is the emergence problem: how do global properties arise from local rules, and can we engineer them predictably?