A goal is a desired state or condition that guides an agent’s decisions. An agent can assess whether input is good or bad in relation to one of its goals. This gives its actions a direction without treating every input as equally useful. section 4.1.2
In conventional problem-solving, a goal state is the final state that counts as a solution. Starting from an initial state, a system considers allowed changes and looks for a sequence that reaches the goal. It may instead choose the next state that appears closest to the goal under a defined distance measure. section 4.1
Desirable and undesirable states
In the AGORA smart-agent language, stabilize marks an acceptable, desirable state for an agent, while unstabilize marks a situation that should never be allowed. These terms make the difference between a preferred condition and one to avoid concrete. section 4.1.2
A goal does not guarantee success. Exploring every allowed sequence can become impractical when possibilities multiply; choosing whichever state looks closest can produce a non-optimal result or reach a dead end. A goal identifies what is wanted, while the method for pursuing it determines what can be achieved. section 4.1
