The Sudan Conversation Report
The Framing Gap in the Sudan war conversation: violence read along two dimensions, as narrated and as landed, plus composition, CAMEO/Goldstein intensity, and actor attribution.
On the narrative reading, which follows each post's own grammar, civilians are the named target of 18% of violence in the Sudan war conversation for 01 Apr 2026 – 30 Jun 2026. On the impact reading, which follows where the harm lands (hospitals, markets, camps, casualty clauses), they are the target of 71%. That 53.1-point Framing Gap means the war is narrated army-vs-army while the harm falls on non-combatants: 84% of civilian-directed violence never names civilians as its object.
Civilian share of target-identified violence is 18% on the narrative reading (each post's own grammar) and 71% on the impact reading (where the harm lands). The distance between the two dimensions measures how much civilian harm is narrated as combatant-vs-combatant engagement.
Coded on the platform's topic categories, 44% of the Sudan war conversation is politics & diplomacy. Only 32% of posts report an actual conflict event; the rest is commentary, humanitarian appeal, politics, and everyday diaspora life. Anyone reading this feed as a live battle map is reading mostly signal-adjacent noise. The war is a minority of the traffic.
Every violence event is read along two dimensions. The narrative reading takes each post's own grammar, who the sentence says did what to whom, and there civilians are the named object of just 18% of target-identified violence. The impact reading follows where the harm lands: attacks on hospitals, markets, camps and schools, and casualty clauses naming children, families and the displaced. There the figure is 71%. The 53.1-point distance between the two is the Framing Gap, and it is a property of the discourse itself: the victims are present in the posts, but as subordinate clauses attached to combatant stories. 84% of civilian-directed violence never names civilians as its object at all. The narrative dimension shows a war between two armies; the impact dimension shows where it actually lands.
The narrative dimension follows each post’s own grammar. The impact dimension follows the harm: civilians are never the perpetrator of coded violence; strikes on hospitals, markets, camps and schools attach to their civilian targets; “UN says…” attributes to the belligerent named, not the reporting agency. Method stated in full below.
Isolating the 373 conflict-event reports and weighting by engagement, the Goldstein reading is -6.0: sharply conflictual. Material conflict (assault, armed clashes, mass violence) is 82% of those events, against 9% coding as diplomacy, mediation, or aid. Kinetic reports also carry the most engagement, so what circulates is more violent than the event mix alone.
Weekly tone of the conflict signal moved -0.6 points from the opening third of the window to the close; the trajectory is escalating. The most intense week was Apr 27 at -10.0. Treat spikes as event-driven and thin weeks as lower-confidence collection rather than calm.
Flagged week: Apr 27. Lower is more conflictual; casualty context pulls kinetic weeks toward −10.
On the impact ledger the heaviest dyad is RSF → Civilians, with subjects on record being SAF, RSF, Civilians and Foreign/Regional. Civilian exposure, at 71% of target-identified violence, is the line to watch: it separates combatant fighting from harm to non-combatants, and it is the number the narrative dimension systematically understates.
| subject ↓ / object → | RSF | Civilians | Foreign/Regional | SAF |
|---|---|---|---|---|
| SAF | 94 | 216.1 | 28 | · |
| RSF | · | 320 | 60.5 | 55 |
| Civilians | 6.5 | · | 6.5 | · |
| Foreign/Regional | 96 | 185 | · | 24 |
Summed casualty-adjusted Goldstein magnitude on the impact reading: violence attributes to where it lands. Civilians are never the perpetrator of a coded violence event, and victim-context events attach to their civilian targets.
- 01Every post is bucketed by the platform’s topic categories into a conversation-composition read (conflict, humanitarian, politics, economy, culture/daily life), so the conflict signal is separated from the surrounding traffic.
- 02Posts that report an action, whether via event verbs (English and Arabic), event-bearing categories, or a violence flag, are coded to a CAMEO root event class carrying a Goldstein −10…+10 weight; commentary and culture are excluded from the intensity read.
- 03Intensity is ordinal, not fixed: casualty context deepens conflict weights toward −10 (after Stoehr et al., ACL 2023), and subject→object attribution adds the who-to-whom dimension the Goldstein scale omits.
- 04Violence is read along two dimensions. The narrative reading takes each post’s own grammar (first and second actor mention, who the sentence says did what to whom) and measures how the discourse frames violence. The impact reading attributes each event to where the harm lands: civilians are never the perpetrator of a coded violence event; attacks on hospitals, markets, camps and schools, and casualty clauses naming children, families and the displaced (in English and Arabic), attach to their civilian targets; and in “UN says…” posts agency attaches to the belligerent the report names, not the reporting source. The distance between the two dimensions is published as the Framing Gap; the impact figures drive the dyad matrix.
- 05The Conflict Intensity Index is the engagement-weighted mean adjusted weight over conflict events; escalation compares the late-window tone to the early-window tone.
Social Knowing, "The Sudan Conversation Report" (Edition 01), 2026. https://socialknowing.com/reports/sudan-conversation-q2-2026
This is a fixed, dated reading; the event coding and every figure are reproducible from the source feed. The situation keeps moving; open the live instrument for the current reading and per-actor feeds.
open Sudan Watch ↗