Overview
The constraint engine decides what a system must not do, and what it must always do. It turns ethical commitments into explicit rules the system checks before it acts, rather than filters bolted on afterwards.
Every proposed action passes through it. The engine either approves the action, adjusts it to fit, or refuses it and says which constraint was at stake.
Hard Constraints
Hard constraints are duties that hold regardless of outcome: lines the system will not cross even when crossing them looks useful. They come from deontology, which treats some obligations as binding in themselves.
Because they never bend, hard constraints are kept few, precise and easy to audit.
Soft Constraints
Soft constraints are preferences weighed against each other: fairness, privacy, cost, the wishes of the people involved. They come from consequentialism, which judges an action by its results.
The engine scores candidate actions against these weights and favours the one with the best overall balance.
Resolving Conflicts
Constraints sometimes pull in different directions. When they do, hard constraints win, then the engine looks for the action a careful, fair-minded agent would choose, the question virtue ethics asks.
Each resolution is logged with its reasons, so people can review the trade-off and refine the rules over time.