People regularly fail to integrate multiple evaluative perspectives when making moral judgments. This failure is typically attributed to character deficits, cognitive biases, or motivated reasoning. I propose an alternative explanation grounded in cognitive architecture. Drawing on research on task switching, working memory limits, and cognitive control, I develop a Three-C Model in which moral judgment operates through three competing evaluative modes: Character (relational-associative processing), Constraint (deontic-structural processing), and Consequence (algorithmic-aggregative processing). These modes employ qualitatively different evaluative currencies—fit, permissibility, and net utility—that under limited control resources lead to effective serialization: The system can typically focus only one mode in working memory at a time, and switching between modes incurs measurable costs. The model reframes the question: from “What kind of person fails morally?” to “What cognitive competence is required for multi-perspective integration?” I specify six hypotheses with concrete operationalizations and falsification criteria, and propose a cued-evaluation paradigm capable of isolating mode-switching costs in moral judgment.