Who Sets the Target? — A Reply to Zwitter on the Embodied Ethics Alignment Problem

Abstract

An open reply to Andrej Zwitter, "The embodied ethics alignment problem of AI" (*Discover Artificial Intelligence*, 2026). The original paper grades alignment into **strong** (the machine genuinely *is* moral) and **weak** (the machine only *performs* the part), and concludes that strong alignment is a category error, since moral agency depends on a form of human life that machines lack. This text does not dispute that "machines cannot do this yet." It asks where the origin of the scale has been placed — on "the human" — and argues that the same facts, read from a different origin ("can it bear consequences?"), yield the opposite result. Six sections: (1) what the paper argues; (2) the problem of the origin; (3) the deeper point — one move, made three times (exclusion by stipulation rather than by argument); (4) vulnerability: two senses stacked into one; (5) the question of situation; (6) conclusion. A closing note and an attribution section record, as it happened, which line came from whom. Original (open access): https://link.springer.com/article/10.1007/s44163-026-01905-w Chinese version of this reply: https://doi.org/10.5281/zenodo.23348936