Reproduce “A Formalization of Kant's Second Formulation of the Categorical Imperative” on OpenResearch

About this paper

Felix Lindner, Martin Mose Bentzen

Lindner and Bentzen developed a formalization and computational implementation of Kant's second formulation of the categorical imperative, enabling artificial agents to evaluate actions based on whether humanity is treated as an end, never merely as a means. Their framework, integrated into the HERA software library, demonstrated its ability to process various moral dilemmas, including strict and wide duties, and showed the distinction between impermissible actions like suicide and permissible self-harm.

Key insights

• The intricate ethical principle of treating humanity "never merely as a means but always at the same time as an end" can be formalized through explicit definitions of "treated as an end" and "treated as a means" using causal agency models.
• The rigorous evaluation of an action's permissibility requires checking if any patient treated as a means is also simultaneously treated as an end, or if any goal negatively affects a patient.
• An additional "Meritorious Principle" is necessary to capture Kant's wider duties, guiding agents to choose actions that positively affect the most patients when multiple options are otherwise permissible.

Summary by alphaXiv.