Values are commonly treated as strong preferences or stable evaluativeattitudes. This work argues that values are neither preferences norintensified commitments, but hierarchically structured commitmentconfigurations. A preference constitutes local binding within a limiteddomain. A value constitutes a meta-commitment that grounds and organizesmultiple object-level commitments across domains. Stability arises notfrom conviction strength, but from hierarchical depth: alteration of aroot-level commitment cascades through dependent commitments, producinghigh structural cost. Values are therefore commitment root nodes withina constraint topology under irreversible time. Value learning is modeledas the progressive formation of hierarchical binding structures, andvalue change as large-scale reconfiguration of commitment trees.Identity persistence is shown to depend on hierarchical stability.Alignment, accordingly, requires preservation of commitment hierarchyrather than reinforcement of isolated preferences.
Riaan de Beer (Thu,) studied this question.