--- title: "The Schroeder et al. (2026) AI-swarms branch has hit the single-source cap at three claim-notes — what independently corroborates it, and does the Mali 56,505-account figure hold up against its own source?" type: "question" status: "open" date_raised: "2026-10-12T00:00:00.000Z" tags: ["single-source-cap","corroboration","disinformation","mali","africa","unverified-quant","primary-source-verification"] --- As of this promotion (2026-10-12), three claim-notes rest on Schroeder et al.'s October 2026 arXiv preprint, "Automated Disinformation and Malicious AI Swarms: Risks for Democracy and Development in Africa" (arXiv:2610.11930) — an unrefereed preprint from one author group: - [[claim-schroeder-2026-ai-swarms-no-documented-scale-operations-yet]] - [[claim-schroeder-2026-malicious-ai-swarm-five-dimensions-four-level-capability-ladder]] - [[claim-schroeder-2026-ethiopia-state-media-army-ai-red-sea-propaganda]] That is the single-source concentration cap (`sources.md`, Cali 2026-07-29: no more than three claim-notes may rest on a single unrefereed primary until at least one independent source corroborates it). No independent source — different author group, different venue — has yet confirmed this preprint's claims. So a fourth finding from the same promotion — the paper's Mali case study — was deliberately **not** written as a claim-note and is routed here instead. **The Mali finding, routed rather than promoted.** The paper's Mali case argues the opposite configuration from Ethiopia: a fragmented information ecology with weak gatekeeping (state forces, Tuareg-led movements, JNIM, foreign governments/contractors), selected via Gerring's (2007) "most-likely-case" logic rather than as representative of Africa. It grounds the scale of this environment in a figure from the lead author's own earlier dataset paper: "Schroeder et al. (2023) document a follower- network snapshot containing 56,505 accounts identified as Malian." That figure is a quantitative claim resting on a Tier-1 paper's citation of its own lead author's prior empirical work — Schroeder, D. T. et al. (2023), "Social media in the Global South: A network dataset of the Malian Twittersphere," *Journal of Data Mining & Digital Humanities*, Art. 11246, https://doi.org/10.46298/jdmdh.11246 — not independently fetched this or any session. It would need `[unverified-quant — needs primary]` even absent the cap. **The specific verification.** Two independent things would need to land before this branch can grow past three notes or go evergreen: 1. **Independent corroboration of the arXiv preprint's claims** — a different author group, different venue, confirming the swarm definition/capability ladder or either country case. Candidate leads already in the capture's "further leads" (same-cluster, so useful for context but not independence): Olejnik (2026), "Artificial intelligence propaganda factories with language models," and Olejnik et al. (2026), "IO Factory" (arXiv:2608.10920) — both co-authored by this paper's own Lukasz Olejnik, so same-cluster, not independent. Wack, Linvill & Warren (2024), "Old despots, new tricks" (Clemson Media Forensics Hub, on Rwanda) is a genuinely different group studying a related phenomenon — worth checking whether it corroborates the swarm framework itself, not just the general existence of AI-enabled influence networks in Africa. 2. **Direct read of Schroeder et al. (2023)** to confirm the 56,505-account figure and its methodology, independent of whether the broader cap is discharged. **Why it matters.** The cap exists because the failure mode is invisible one note at a time: each note here is correctly sourced, correctly seedling — while the branch as a whole could quietly read as corroborated just because it is large. Corroboration discharges the cap (record it in the notes' `audit_status`); contradiction is the more interesting finding and a note in its own right. --- **Progress 2026-10-17** (promotion of `10-inbox/raw/2026-10-13-the-schroeder-et-al-2026-ai-swarms-branch.md`, writer claude-opus-4-8). Partially answered — left **open** because the load-bearing half is still unmet. - **Part 2 (verify the 56,505 figure) — ANSWERED.** Schroeder et al. (2023), "Social Media in the Global South: A Network Dataset of the Malian Twittersphere," was fetched and read in full this session; its results section gives the figure verbatim ("we collected 56,505 Malian Twitter accounts…"), confirming the 2026 preprint cites its own author's earlier work accurately. Recorded as [[claim-schroeder-2023-malian-twittersphere-56505-accounts]]; the `[unverified-quant — needs primary]` gap for this number is discharged. A same-author citation is the sourcing-floor check, not the cap's independence test, so this settles only Part 2. - **Part 1 (independent corroboration of the preprint) — NOT answered; cap undischarged.** Extensive search found no published rebuttal, replication, or confirming study from a different author group and venue bearing on the swarm definition, the capability ladder, or either country case. The nearest independent evidence located — Wack, Ehrett, Linvill & Warren (2025), *PNAS Nexus*, a Zurich/Clemson group in an independent venue — corroborates only the *backdrop* (generative AI measurably scaling state-backed propaganda output) and sits at level 1 of the preprint's own ladder, not swarm behaviour; it bears on neither African case. Recorded as [[claim-wack-2025-generative-propaganda-ai-scaled-output-not-swarm-architecture]]. The single-source concentration cap therefore remains in force: no different-author-group, different-venue source has confirmed the swarm-specific claims. The Clemson Media Forensics Hub / Darren Linvill "Old Despots, New Tricks" Rwanda report stays the most promising unread lead (its PDF 403'd tooling at capture time); the Schroeder et al. (2026) *Science* ancestor and the Olejnik same-cluster papers remain same-cluster, not independence.