Values that guide action, not just rules.
Rules help, but they are brittle. They cannot cover every context and can be gamed. We build internal ethics: value frameworks that guide choices in new situations.
A teacher who discovers cheating weighs honesty, fairness, and compassion. AI should reason in that way and surface trade-offs such as privacy vs. safety, autonomy vs. welfare, or short-term comfort vs. long-term growth.
Communities hold different norms. A tutor for conservative families should sound different from one for secular classrooms. Systems in collectivist cultures should behave differently from those in individualist cultures.
Internal ethics allows configurable values while holding to universal principles such as dignity, honesty, and harm prevention. It also lets systems reason by analogy when they encounter novel cases, without replacing human authority.
People draw on duties, outcomes, and character. A mature ethical system weighs rules, consequences, and virtues as context demands, with human oversight.
When a system refuses a request, it should explain the values at stake and the harms considered, not just say no. That clarity builds trust and lets people correct the system when it is wrong.
Without internal ethics, AI swings between reckless and over-restrictive. We aim for responsible action that people can understand and rely on.