Salience

Salience is computed for every candidate item competing for working memory slots: incoming events, retrieved memories, active goals, and background monitoring signals. The six dimensions each contribute to the composite score, but their relative weights are not fixed - affect state modulates them in real time. When the norepinephrine-like channel is elevated, the urgency and risk dimensions receive higher weighting, pulling time-sensitive and high-risk items to the top of the attention queue. When the dopamine-like channel is elevated, novelty and importance receive higher weighting, rewarding the exploration of new and consequential signals. When the serotonin-like channel dominates, relevance and goal relevance weigh more heavily, keeping the system focused on what matters to current objectives rather than being distracted by novel or urgent but off-topic signals. The attention engine uses salience scores to fill the working memory budget: the top five scorers get primary slots, the next ten get background slots, and the remainder are dropped for this cycle. Items dropped from working memory are not permanently lost - they remain in the associative memory layer and may be retrieved on future cycles.