Formalizing Kantian Ethics: Formula of the Universal Law Logic (FULL)
New logic system enables AI to evaluate actions based on purpose, not just pre-programmed rules.
Researcher Taylor Olson has published a novel paper, "Formalizing Kantian Ethics: Formula of the Universal Law Logic (FULL)," proposing a new framework for machine ethics. The work addresses two key limitations in current approaches to building Artificial Moral Agents (AMAs): the failure to consider an agent's purpose when acting, and the reliance on enumerating a finite set of human moral intuitions as axioms. Instead of hard-coding rules like "do not harm," Olson's system formalizes Kant's Formula of the Universal Law (FUL), creating a logic where an action is moral only if the agent can will its underlying maxim to become a universal law.
The FULL system is a multi-sorted quantified modal logic that formalizes concepts like causality, agency, and universalizability. In tests on three classic cases from Kantian ethics, the logic demonstrated an ability to evaluate the morality of agents' actions for specific purposes without any built-in moral axioms. It only requires sufficient non-normative background knowledge about the world. This represents a significant shift from rule-based ethical systems toward procedural, reasoning-based ones, moving AI closer to autonomous moral reasoning rather than simple rule-following.
- Formalizes Kant's categorical imperative into a multi-sorted quantified modal logic called FULL.
- Allows AI to evaluate actions based on purpose and universalizability, not pre-set rules.
- Demonstrated on three Kantian ethics cases using only non-normative background knowledge.
Why It Matters
Paves the way for more autonomous, robust AI agents capable of principled moral reasoning in novel situations.