Beyond Classical Logic: Aberration Logic, Evaluative Dimensions, and the Transition from AI Ethics to Evaluative AI

Contemporary discussions of artificial intelligence increasingly focus on AI Ethics, AI Safety, Responsible AI, Explainable AI (XAI), and alignment. While these approaches address important aspects of artificial-system behavior, they often lack a unified framework capable of representing multiple evaluative perspectives simultaneously. Building upon the Ethics-by-World-Notions approach introduced in Beyond Evaluation (Gamboa 2026b), this paper develops a formal framework for representing, comparing, and communicating evaluative structures within artificial systems. The paper introduces Aberration Logic as a deviation-sensitive logical framework that extends classical truth-based assessment by enabling the representation of graded evaluative deviations. On this foundation, major truth theories are reinterpreted as complementary Truth Dimensions rather than competing theories of truth. Four principal Truth Dimensions are developed: Veritas, Telos, Ethica, and Aesthetica, corresponding respectively to correspondence-based, pragmatic, ethical, and aesthetic modes of evaluation. These dimensions enable multidimensional evaluative representations through evaluative vectors and configurable evaluative spaces. The framework further develops Evaluative AI as an extension of conventional AI architectures through the introduction of explicit evaluative representations based on Truth Dimensions, World Notions, and Multi-Valued Aberration Logic. Both Truth Dimensions and World Notions are treated as configurable components, allowing evaluative systems to adapt to different domains, stakeholders, and evaluative objectives while maintaining transparency and comparability. In addition, the paper introduces evaluative distance, trust, and believability as emerging concepts grounded in multidimensional evaluative representations and their systematic comparison. More generally, the framework supports assessment, comparison, simulation, explanation, and decision support across alternative evaluative perspectives. The transition from AI Ethics to Evaluative AI is therefore interpreted not as a replacement of ethical evaluation, but as its integration into a broader evaluative framework capable of supporting transparent, configurable, and human-centered forms of artificial evaluation. Version note: This is Preprint v2. Compared with Preprint v1, the formal definition of the AND operator in Aberration Logic has been revised from a product-based aggregation rule to the minimum rule. The corresponding examples, figures, and appendix sections have been updated accordingly. This version also adds recent references on Responsible AI, AI alignment, AI risk management, and AI regulation.