Abstract
Building on Nunes Filho's recent positioning of paraconsistent logic as a constitutive element of Latin American philosophy (RUDN Journal of Philosophy, 2025), this paper traces a continuation of that tradition into the contemporary problem of artificial intelligence alignment. We argue that the Latin American paraconsistent project — initiated by Miro Quesada's coining of the term in 1976 and formalized by Newton da Costa's C-systems — finds its natural twenty-first century extension in Florentin Smarandache's neutrosophic logic (1995), which generalizes paraconsistency by introducing three independent dimensions of evidence: truth (T), indeterminacy (I), and falsity (F), with no sum constraint. We further contextualize this development within the broader Eastern European tradition of non-classical logic, from Vasiliev's non-Aristotelian logic (1911) and the Romanian school of Lukasiewicz–Moisil multi-valued algebras to Smarandache's emigration from Romania in 1988 and his subsequent founding of neutrosophy. We then show that contemporary AI alignment frameworks — Reinforcement Learning from Human Feedback, Constitutional AI, Direct Preference Optimization — operate under an implicit epistemological monism that mirrors precisely the classical-logic universalism the Latin American tradition critiqued. Anthropic's own documentation of Collective Constitutional AI confirms that contradictions among diverse constitutional principles are resolved by majority vote — what we term enforced convergence rather than alignment. We propose Pluriversal Alignment: a research program that operationalizes paraconsistent neutrosophic triplets as reward signals, constitutional evaluations, and policy outputs, allowing T+F>1 as a structural feature rather than an error. We close by arguing that the Latin American paraconsistent tradition, in convergence with its Russian and Eastern European counterparts, is uniquely positioned to contribute the formal apparatus that the decolonial critique of AI has named philosophically but not yet formalized.