A goal is a desired outcome that guides behavior. Human goals can arise from needs, preferences, commitments, and social pressures; they may persist, shift, or dissolve as circumstances change. Artificial neural networks, by contrast, do not generate goals for themselves: their objectives, inputs, and stopping conditions come from a larger system. Behavior that appears purposeful reflects the task being served, not an internal orientation toward an outcome. section 5.4.2 section 5.4.2
Distinguish goal-generating agency from behavior directed toward a goal:
- People can decide what to pursue, what to avoid, and when to redirect effort. section 5.4.2
- A model can produce coherent, adaptive-looking behavior within a defined task without deciding what matters or revising its own objective. section 5.4.2
A goal state in a puzzle
In Tower of Hanoi, Checker Jumping, River Crossing, and Blocks World, a puzzle starts in an initial state and sets rules for legal moves. The goal state is the specified outcome to reach without violating those rules; solving requires a valid sequence of steps, not just a final answer. section 9.4.1













