Post
A flowchart for NPC brains that makes them look smart without actually being smart.
Behavior trees are hierarchical structures that organize AI decision-making into a tree of nodes that are evaluated from top to bottom. The tree contains action nodes (do something), condition nodes (check something), and composite nodes that control flow -- sequences (do all children in order, fail if any fails), selectors (try children until one succeeds), and decorators (modify child behavior). Each tick, the tree is traversed and nodes return success, failure, or running. The beauty of behavior trees over finite state machines is modularity: subtrees can be reused, debugged independently, and combined without the spaghetti of state transition logic. Most modern game AI -- from Halo's enemies to the companions in The Last of Us -- runs on behavior trees.
Example
Halo 2 pioneered the use of behavior trees for its enemy AI, allowing Covenant soldiers to dynamically choose between flanking, retreating, using cover, throwing grenades, and calling for backup based on the evolving combat situation. The system was so effective that Halo's AI is still considered a benchmark over two decades later.
Why it matters
Behavior trees made complex, believable AI practical for game developers. They provide a visual, modular, and debuggable way to build NPC intelligence that scales from simple patrol-and-attack routines to sophisticated tactical decision-making without drowning developers in spaghetti logic.
Related concepts