Do We Even Need the Word "Value"? — Generating Ethical Weight from Action Differences, Downstream Possibilities, and Recoverability in AI

This paper asks whether ethical theory requires the primitive concept of value, or whether ethically significant weighting can be reconstructed from differences in action, downstream possibility, harm propagation, and recoverability. Rather than beginning with intrinsic value, goodness, or a fixed hierarchy of ends, the analysis tracks how alternative actions alter reachable futures, constrain later options, distribute burdens, and change the possibility of repair. Ethical weight is generated when these differences become non-redundant for subsequent action and cannot be collapsed into mere preference or prediction. Recoverability plays a central role: two outcomes with similar immediate costs may differ ethically when one preserves future correction, reversal, compensation, or reconfiguration while the other closes those possibilities. The framework is then applied to artificial intelligence, where autonomous and semi-autonomous systems can produce downstream effects that are distributed across agents, institutions, infrastructures, and time. The paper distinguishes ethical weighting from the claim that value exists as an independent property, and asks whether value vocabulary remains necessary once the relevant action-differences and downstream constraints have been operationally specified. The broader claim is that some ethical work may be performed by configuration-sensitive weighting of consequences, responsibilities, and recoverability without presupposing a primitive ontology of value, while leaving open whether the concept of value becomes non-redundant at higher-order normative levels.