KVA Research · May 2026 · Conference Paper (NLPAICS 2026)

By Maryam Fooladi, Federico Bottino and Alberto Trivero

5 pages · 12 min read

Bloc-Conditional Event States: Measuring Cross-Coverage Divergence for Threat-Intelligence Analysis

A Reuters dispatch and a PressTV dispatch on the same event both pass every source-level check open-source intelligence relies on. What separates them is how the event is framed. This paper asks whether that difference can be measured — and shows that the divergence between state-aligned and mainstream-Western coverage runs far deeper than the polarization inside any single domestic press.

The blind spot in source-level intelligence

Open-source intelligence routinely characterizes news coverage through source-level signals: domain registration, linking graphs, account coordination, outlet reputation. These signals are effective against fabricated infrastructure. They are far less informative about content from established state media, which is correctly attributed and passes source-level heuristics without difficulty.

The paper poses a comparative question: does aggregated framing from one bloc of outlets diverge measurably from another, and does cross-bloc divergence exceed the editorial polarization already present within a single domestic press tradition? The contribution is measurement, not a detector — the work constructs the observable and characterizes its behavior rather than calibrating thresholds for classification.

Framing as a density matrix

Building on the event-state density-matrix formalism of Bottino et al. (2026), the method partitions outlets into four editorially-coherent blocs — mainstream-Western, US-right, US-left, and state-aligned. Each article maps to a vector on a 15-dimensional framing-feature space, anchored on the canonical Semetko & Valkenburg news frames (responsibility, morality, economic consequences, human interest, conflict). Each bloc is then represented as a density matrix, and divergence between any two blocs is the trace distance between their matrices.

The target observable is the distance between state-aligned and mainstream-Western coverage, benchmarked against the distance between US-right and US-left.

What the two events show

The method is applied to two contested events: the Hormuz blockade of 2026, with 16 directly state-aligned outlets, and the death of Navalny in 2024, with a 14-outlet state-aligned bloc. On both events, the cross-bloc distance between state-aligned and mainstream-Western coverage exceeds the US-right/US-left distance by a factor of roughly 1.8.

Decomposing the state-versus-mainstream difference along its top eigenvector recovers interpretable, event-appropriate framing axes: economic-consequences framing on Hormuz, and morality framing on Navalny. The gap is not a single opaque score — it resolves into the specific dimensions along which the blocs tell different stories.

Measurement, not detection

The work stops short of building a classifier. At both events it demonstrates the construction and characterizes what the observable does, leaving threshold calibration and downstream threat-intelligence applications — coverage that lacks content-level analogues — as the next step. The result is a content-level signal that complements source-level OSINT exactly where source-level OSINT goes quiet.

KVA Research · Kakashi Ventures Accelerator Srl · Turin, Italy
Published May 2026