Credence First: Against Plantinga's Warrant as an Epistemic Upgrade

Alvin Plantinga's warrant program is arguably the most sophisticated attempt in contemporary epistemology to identify the property that, when added to true belief, yields knowledge. This paper argues that the project fails, not because the analysis of proper function is incorrect, but because "knowledge" itself is a conceptual hallucination that actively interferes with epistemic clarity. We advance an eliminativist thesis: "knowledge" is a coarse-grained folk artifact that should be discarded in serious epistemology in favor of credencing -the active, continuous modulation of belief in response to evidence. Through a formal demonstration using strictly proper scoring rules, we prove that a cognitive design plan aimed at maximizing a binary status like warrant must, in specific evidential conditions, minimize continuous accuracy. Consequently, warrant functions not as an epistemic upgrade, but as a biasing incentive that encourages miscalibration. We apply this critique to the internal logic of the sensus divinitatis , arguing that it relies on a "calibration vacuum" where unjustified certainty is mistakenly labeled as knowledge.