Decision Epistemic Strength: A Credibility Weighted Framework for Decision Governance Under Uncertainty

Decision-making in complex systems often requires converting heterogeneous, noisy, and partially conflicting evidence into actionable choices under epistemic uncertainty. Although statistical inference and predictive modeling quantify uncertainty and prediction error, they do not directly quantify the epistemic legitimacy or dominance of competing actions after evidence credibility is considered. This Method Article introduces Decision Epistemic Strength (DES), a deterministic, credibility-weighted metric designed to govern decisions explicitly in decision space rather than hypothesis-testing space. The DES framework follows a four-layer pipeline: (i) epistemic zoning of evidence using standard-error-based or reliability-defined partitions, (ii) local characterization within zones using conservative stability-aware (iii) zone-specific decision assignment using predefined decision rules, and (iv) credibility-weighted fusion of decision votes followed by dominance quantification. For each candidate action d in D, a cumulative credibility score is computed as S_d = sum_{i: d_i = d} w_{z_i}, and DES is formally defined as the normalized dominance margin between the winning action and the strongest competitor: DES = (S_win - S_lose) / W_total, where W_total = sum_{d in D} S_d, yielding a bounded score in [0,1] that supports graded governance states (Decisive, Conditional, Deferred). Mathematical verification establishes deterministic reproducibility, boundedness, monotonicity, and robustness to weak-signal dominance. The method is demonstrated using simulated clinical and epidemic policy scenarios and illustrative real-data governance examples (ICU monitoring and clinical AI), showing coherent behavior under evidence conflict, perturbation, and credibility variation. DES complements inferential and predictive methods by providing a transparent, auditable governance layer for accountable decision authorization under epistemic uncertainty.