Utility AI
A decision model where every possible action is scored by a utility function and the agent picks the highest — or samples weighted among the top candidates.
Rather than the discrete branches of a behavior tree, utility AI maps each relevant factor — health, distance, ammunition, threat — through a response curve to a number between 0 and 1. Every action's score is recomputed at each decision point and the highest wins. The result is smooth and graded: an agent switches goals as circumstances cross a threshold, without anyone having authored that particular transition.
The needs system in The Sims (2000) is the best-known example, and Dave Mark formalised the approach as the "Infinite Axis Utility System" across a series of GDC talks. The strength is scalability — new actions are added without touching the old ones — but the curves need careful tuning, or the agent locks onto choosing the same thing every time.