관계 연산자 JTB 기반 인식론적 러닝 설계 — 딥러닝 + 초월(인식론적 사고) 러닝을 위한 JTB-GEL 프레임워크

This article is a study that realizes 'know' as a true value judgment of JTB (justified true belief), not a simple prediction probability. Propose a framework. User-defined relational operators * (I know), / (I don't know), and L, the being of justification Built into AI, the model (1) belief (belief) score b, (2) evidence-based justification score ℓ, (3) truth t through verification Separate learning and reasoning. In the proposed JTB-GEL (JTB-Gated Epistemic Learning), the standard Position deep learning (expression learning and prediction) as faith B learning, and justify transcendent (emological thinking) learning L's creation and evaluation, Defeater detection, defined as a meta-roup that learns the choice of verification behavior. Also, when declaring knowledge, 'the basis Propose a knowledge certificate format that outputs the evidence bundle together, and explain It fixes the possibility of possibility and revery with system rules.