Topinka, R. (2026). Feed them slop: Alignment, attunement, and the affective lure of slopaganda. European Journal of Cultural Studies. https://doi.org/10.1177/13675494261468632
Summary
This article theorizes “slopaganda” (or “agitslop”) — AI-generated political content that is obviously fake yet affectively resonant — as a phenomenon distinct from deepfakes. Where deepfakes depend on deception, slopaganda works as an affective lure that hooks attention (even negative or dismissive attention) within the moods and affordances of social media feeds. Topinka argues that AI models, through training, fine-tuning, and alignment (especially RLHF), become “ideological attunement machines” optimized to sense what resonates in their sociotechnical environments. The result is a form of political content that intensifies rather than disrupts existing ideological formations — racism, White nationalism, engagement-driven platforms — without ever making coherent political claims.
Key Contributions
- Reframes slopaganda from a problem of deception/belief manipulation to one of affective attunement and the lure of attention.
- Introduces the concept of AI systems as “ideological attunement machines” optimized for feeling, relevance, and response.
- Bridges rhetorical theory (ambient rhetoric) with technical AI processes — fine-tuning, RLHF, alignment, and prompting — showing their productive resonance.
- Offers an analytical lens for why crude, meaningless, or incoherent AI content nonetheless resonates and coordinates ideology.
- Extends the critique of AI beyond bias toward how models and their prompters “feel out” ambient environments to produce resonant outputs.
Methods
Conceptual and theoretical analysis grounded in cultural studies, rhetoric, and affect theory. Topinka performs close readings of contemporary examples of AI political content (the “Trump Gaza” video, the “Kirkinator” videos, AI Blackface slop, Trump administration slop-posts, an Iranian anti-Trump music video), synthesizing concepts including Rickert’s ambient rhetoric, Sampson’s affective lure, Massumi’s affect/emotion distinction, Noordenbos and Tuters’ ambient propaganda, and Burkhardt and Rieder’s “foundation models as platform models.” The analysis also attends to the political economy and labor of alignment — RLHF, exploited Global South workers, and WEIRD-dominated image labeling.
Findings
- The “Trump Gaza” video circulated through emotional resonance and outrage rather than by persuading anyone of a coherent claim — slop as affective coagulation, not belief manipulation.
- AI slop overwhelmingly targets women and racial others (immigrants, Black protestors, Black lawmakers), giving form to reactionary fantasies.
- Image and video generators reproduce hegemonic norms because training/labeling is dominated by WEIRD-country workers (~95%) and caters to existing consumer expectations.
- The “Kirkinator” videos are ideological (invoking White nationalist imagery like Agartha) yet incoherent, circulating even on left-wing meme accounts precisely because they refuse representational and ideological coherence.
- Ambient attunement to internet slop can frame real-world violence, as with the “for Agartha” inscription in the Jakarta mosque attack.
- The prompt itself functions as a key site of rhetorical attunement, with foundation models having no intrinsic function until prompted by a rhetor attuned to ambient affects.
- Contra Bernays, slop is disordered and directionless — coordinating affect and ideological orientation without bringing “order out of chaos.”
Connections
This paper’s affective, resonance-centered account of AI political content offers a counterpoint to work measuring the persuasive or belief-changing power of AI-generated messaging, such as Hackenburg2025-dj and Hackenburg2026-ud. Its treatment of AI “slop” as a distinct media object connects to broader generative-AI media critiques like Gerard2025-br and to analyses of reactionary and memetic online politics found in Marwick2026-qd and Copland2025-em.
Podcast
A research-radio episode discusses this paper: 🎧 MP3 · Spotify