Cognitive Architectures for Agents

Goal

Structure agent reasoning, memory, and decision-making with patterns from classic cognitive architectures (SOAR, ACT-R, LIDA) — beyond ad-hoc prompting.

1. SOAR — Problem Space + Operators

Break "think" into a proposal-evaluation cycle rather than one step: Elaborate (infer from state) → Propose candidate operators → Evaluate by heuristics → Select best → Apply to change state.

def cognitive_cycle(state):
    state = enrich_context(state)                 # 1. elaborate
    options = generate_candidates(state)          # 2. propose
    scored = evaluate_candidates(options, state.goal)  # 3. evaluate
    return apply_operator(state, select_winner(scored))  # 4-5. select + apply