Skip to main content
Correspondia
Paper overview

Aggregation and Composite Index Principles

The default aggregation principle is profile first. Because Organizational Correspondence is relational, multi-level, temporal, evidential, and often non-compensatory, heterogeneous results should remain visible unless aggregation preserves construct meaning, level identity, evidence warrant, uncertainty, temporal context, and interpretation boundaries.

Figure 4 shows the profile-first aggregation logic.

Figure 4

Profile-First Aggregation and Non-Compensability

FIG-0019 depicts Profile-First Aggregation and Non-Compensability. It represents Profile, Admissible aggregation, Provisional aggregation, Prohibited aggregation, Compatibility, Weighting, Compensability, Bottleneck, Uncertainty, Coverage, Critical failure, Validation, Governance review. The intended relationships are: Profile preservation is default; aggregation passes through compatibility, propagation, compensability, bottleneck, validation, and governance gates. The figure should be read with this boundary: Aggregation is conditional and invalid aggregation is not low correspondence.
Figure 4. Profile-First Aggregation and Non-Compensability. Main Figure 4 shows profile preservation as the default and routes aggregation proposals through compatibility, propagation, compensability, bottleneck, and validation gates. Blocked paths indicate invalid aggregation, not low correspondence.

Aggregation is not prohibited. It is conditional. A bounded aggregate may be appropriate when components measure compatible constructs at compatible levels, share defensible result representations, use comparable evidence regimes, have compatible temporal frames, preserve uncertainty, and pass governance review. But aggregation is not admissible merely because several numbers exist. It must state assumptions about construct compatibility, scale properties, evidence coverage, confidence, uncertainty, missingness, temporal comparability, procedure versions, non-compensability, and validation status.

The most important risk is compensability. Some correspondence failures cannot be offset. A missing evidence chain may not be compensated by strong semantic similarity. A broken dependency in a critical path may not be offset by strong correspondence elsewhere in the network. A stale outcome claim may not be compensated by high traceability from intent to decision. An AI-generated synthesis may not compensate for missing source evidence.

Composite Correspondence Indices are therefore future research objects, not current claims. They may become useful for bounded studies if component validity, weighting, normalization, sensitivity, robustness, invariance, and interpretive boundaries are established. Until then, profiles, paths, configurations, trajectories, and claim-specific indicators are more defensible than universal scores.