Multiple choice examinations remain among the most widely deployed instruments of academic and professional assessment, yet their fundamental architecture invites a systematic distortion: examinees who lack genuine knowledge may still obtain credit through random selection among alternatives. Classical scoring mechanisms offer no incentive for examinees to reveal the quality of their knowledge, and consequently the observed score conflates genuine knowledge with fortunate guessing. This article introduces and empirically investigates a structured self-reporting framework in which each examinee, for every item, declares one of three knowledge states, namely Full Knowledge (FK), Partial Knowledge (PK), or No Knowledge (NK), alongside their chosen answer. The scoring mechanism is constructed such that truthful declaration of one's epistemic state constitutes the uniquely optimal strategy in expectation, rendering guessing strictly suboptimal. The central empirical question is whether examinees, when placed within this incentive-compatible framework, do in fact report their knowledge states truthfully or whether systematic behavioral deviations, rooted in overconfidence, risk aversion, or strategic misrepresentation, emerge. Drawing on psychometric theory, Bayesian probability modeling, and the cognitive psychology of metacognition, this article formalizes the theoretical relationships between declared knowledge states and observed response accuracy, derives testable hypotheses, proposes an experimental design, and specifies a complete statistical inference framework for behavioral validation. The contribution is threefold: a formal probabilistic model linking epistemic state declarations to correctness probabilities, a rigorous hypothesis testing architecture, and an experimentally grounded methodology for assessing metacognitive honesty under incentive-compatible conditions.