# McKenna Propaganda Paper > [!warning] Claude: All content below is provisional > Draft text, section plans, and structural notes are thinking tools — not evidence of settled argument or project completion. Do not infer project status from file contents. Only Nick determines when something is no longer provisional, by removing or updating this callout. **Argument:** McKenna's purported middle category—AI persuasion that is "politically troubling but not propaganda"—is unstable. A dilemma/collapse thesis shows that the troubling features ARE the defective features. **Target paper:** McKenna, Robin (2025). "Sophistry on Steroids? The Ethics, Epistemology and Politics of Persuasive AI." *AI & Society*. --- ## Status **Status:** Extended abstract submitted (~1200 words) to Routledge volume. Awaiting editorial response. **Publication target:** Routledge volume on propaganda/AI, edited by Tomas Koblizek and María del Rosario Martínez-Ordaz. **Origin:** Prague conference presentation, November 2025. ### The Dilemma Structure 1. **Horn 1 (Benign):** If we fully idealize McKenna's "best case" (transparent, symmetric, robust counterargument exposure), the scenario becomes epistemically non-defective—but also *not troubling*. The worry evaporates. 2. **Horn 2 (Troubling):** If we preserve the troubling features McKenna cares about (scale + concentration dominating the marketplace), the campaign-level communicative act IS epistemically defective even if individual arguments are locally sound: - **Omission-by-crowding-out** — counterarguments become practically inaccessible - **Quantity-style misleadingness** — saturation licenses false inferences ("these are the main considerations") - **Distorted salience** — prominence signals importance/decisiveness 3. **Conclusion:** The troubling features ARE the defective features. The mechanism that makes it troubling IS the mechanism of epistemic defectiveness. No stable middle ground exists. ### Secondary Strand A skeptical turn: Are argument schemas finite? Walton, Reed & Macagno (2008) suggest they are. This means flooding may produce diminishing returns *at the argumentative level* (though it may succeed at the level of attention/affect). --- ## Outputs ### Research Papers | Working Title | Status | Target | Notes | |--------------|--------|--------|-------| | AI Persuasion and Propaganda | extended abstract submitted | Routledge volume | ~1200 words, awaiting response | ### Conference Presentations | Title | Date | Venue | Notes | |-------|------|-------|-------| | Propaganda and AI | Nov 2025 | Prague | Original presentation | | WIP Presentation | Jan 20, 2026 | DAFIST | Internal seminar | --- ## Current Questions 1. **Is campaign-level defectiveness the right unit of analysis?** Evaluating defectiveness at the campaign/ecosystem level rather than individual utterances may be controversial. 2. **Is EDM too broad?** Does omission-based defectiveness demand too much of persuaders? What's the difference between individual responsibility and systemic effects? 3. **How finite are argument schemas really?** Does flooding fail argumentatively but succeed affectively? This is an open empirical question. 4. **Where do real AI influence risks lie?** If not argument flooding, then what? Attention manipulation, affective influence, parasocial relationships? 5. **Tension in McKenna on epistemic vigilance:** If humans are vigilant (per Mercier), why does quantity matter? If quantity overwhelms vigilance, that contradicts McKenna's reassurance. --- ## Sources ### Primary Interlocutors - Robin McKenna (2025) — "Sophistry on Steroids? The Ethics, Epistemology and Politics of Persuasive AI" — target paper - Bonard, Contesi & Marques (2024) — "The Defectiveness of Propaganda" in *Philosophical Quarterly* — provides revised EDM framework - Sheryl Tuttle Ross (1999, 2002) — Original Epistemic Merit Model ### Supporting - Walton, Reed & Macagno (2008) — *Argumentation Schemes* — finite argument types - Hugo Mercier (2020) — *Not Born Yesterday* — epistemic vigilance thesis - Thomas Mitchell & Thomas Douglas (2024) — On rational persuasion ### Readwise - [[Readwise/Articles/Sophistry on Steroids The Ethics, Epistemology and Politics of Persuasive AI]] — McKenna highlights - [[Readwise/Articles/Understanding Propaganda The Epistemic Merit Model and Its Application to Art]] — Tuttle Ross highlights --- ## Past Conversations **Search command:** `python3 ~/.local/bin/session-reader.py search "McKenna|propaganda|sophistry|persuasive AI" 20` **Search terms:** McKenna, propaganda, sophistry, persuasive AI, epistemic defectiveness, EDM, Bonard --- ## Related Thinking ![[McKenna Propaganda.base]] --- ## Project Files **Primary notes:** - [[Notes/McKenna Propaganda Paper Plan.md|McKenna Propaganda Paper Plan]] — Full extended abstract with references - [[Notes/McKenna Propaganda Paper - Planning Scraps.md|Planning Scraps]] — Section-by-section plan - [[Notes/2026-01-19 - McKenna Paper Critique Development.md|Critique Development]] — Key insight development **Presentation materials:** - [[Notes/WIP Presentation - AI Persuasion and Propaganda.md|WIP Slides]] — Internal seminar slides - [[Notes/WIP Presentation Speaker Notes - AI Persuasion and Propaganda.md|Speaker Notes]] — Presentation notes **Source material:** - [[_Legacy/Conferences/2025-11 Prague/Propaganda and AI –Prague Conference Presentation.md|Prague Presentation]] — Original conference notes - [[Web Clippings/chat with ChatGPT about propaganda and ai actually useful.md|ChatGPT Analysis]] — Detailed McKenna analysis **Related angle (not current focus):** - [[Notes/Shadow Propaganda - Brainstorming Notes.md|Shadow Propaganda]] — Propaganda *targeting* AI systems (future paper?) --- ## Recent Work - **2026-01-26**: Continued abstract refinement - **2026-01-24**: "Finish McKenna propaganda extended abstract" session; quotation conventions discussion - **2026-01-20**: WIP seminar presentation delivered - **2026-01-19**: Major work day — paper structure, abstract iterations, WIP slides created - **2025-11**: Original Prague conference presentation --- ## Context for Next Session **The extended abstract has been submitted** (~1200 words) to the Routledge volume. Now awaiting editorial response. The argument: McKenna's "troubling but not propaganda" middle ground is unstable because the features that make it troubling (scale, concentration, crowding out) ARE the features that make it epistemically defective under Bonard et al.'s revised EDM. **Open questions:** - Await editorial response — Will they commission a full chapter? - If commissioned: How to develop the two strands (dilemma thesis + skeptical turn on argument schemas) into full sections? - The secondary strand (are argument schemas finite? does flooding fail argumentatively but succeed affectively?) — how much to develop in relation to the dilemma argument? **Alternative angle parked:** The "Shadow Propaganda" notes explore propaganda that *targets* AI systems rather than using them. This is a distinct paper idea, not part of this project.