The Structured Quantitative Assessment Framework (SQAF): An Auditable Multi-Dimensional Computational Architecture for Complex Evidence Evaluation

Abstract

Complex evidential reasoning often involves events whose occurrence depends on several qualitatively different processes and evidence items whose relevance crosses those processes. The difficulty is not that established probability theory is inadequate within a well-defined model. The difficulty arises earlier: the analyst must determine which conditional processes are active, how heterogeneous evidence bears on them, which established computational model is appropriate for each process, and how the resulting outputs can be integrated without erasing their structure. This paper presents the Structured Quantitative Assessment Framework (SQAF) as an auditable, event-activated, multi-dimensional integration architecture. SQAF is not a fixed three-dimensional model and does not propose a substitute for classical probability. It preserves conventional probabilistic, sequential, decision-theoretic, and simulation-based calculations within their appropriate domains, while adding an intermediate vectorized layer for event decomposition, dimension activation, evidence projection, bounded capacity allocation, and cross-dimensional reporting. Ontological, Sequential, and Strategic dimensions are commonly activated interfaces rather than compulsory components. A Generative or other extension may be activated when a lower-dimensional or single-process representation risks omitting, compressing, conflating, or obscuring information relevant to the structure, interpretation, or auditability of the inference. The framework treats dimensions as functionally orthogonal computational spaces: they answer different inferential questions, although the real event may depend on interactions among them. Evidence can therefore be projected onto more than one activated space, but its total directional contribution is bounded through Euclidean normalization and evidential-force conservation. Dimension-local outputs are integrated only through a pre-declared rule and remain visible beside the scalar summary. Projection sensitivity, activation sensitivity, parameter sensitivity, and synthesis sensitivity are incorporated into the robustness layer so that unavoidable expert judgment is converted into explicit, bounded, and challengeable computational assumptions. A staged Monty Hall example demonstrates how event structure activates additional analytical interfaces without changing the underlying probability rules, and the Dinggong pottery-inscription analysis provides a detailed mechanism stress test. SQAF is thus positioned as an organic vectorized integration framework for heterogeneous evidence, not as an alternative probability calculus.