Defective, Amplified, Contested: Folk Conceptualizations of AI Slop

Zhixin Li; Jaime Banks

Home/Archive/arXiv

arXiv · 2610.14231

Defective, Amplified, Contested: Folk Conceptualizations of AI Slop

Zhixin Li·Jaime Banks

View original source ↗Read source document ↗Export metadata ↓

Abstract

AI slop has flooded social media feeds with overprocessed and derivative content--Jesus fusing with a shrimp, a humanoid orange chubby cat shoplifting, and a "hero" dog rescuing a human child after a plane crash. This might be no laughing matter, as waves of media coverage have taken a distinctly negative bent, emphasizing its low quality, ethical concerns, and threats to human creativity. Despite extensive attention from politicians and technologists, the perspectives of everyday users--the very audience most exposed to this phenomenon--remain largely unexplored. This descriptive work explored how Reddit users experience, conceptualize and contest AI slop. Inductive thematic analysis of posts and comments across six slop-related subreddits identified four themes covering constitutional defects, algorithmic overreach, user resistance, and entertainment outcomes--ranging from anxieties about eroding agency, moral norms, creativity, and existential questions to defensive distinctions between quality AI content and slop. We derive a user-driven definition: AI slop is audience-contested AI-generated content characterized by constitutional defects and enabled by platform logics. We operationalize slop through a facet-theoretical framework to inform future research into this fuzzy category.

Explore related subjects

cs.CY cs.HC

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Zhixin Li, Jaime Banks. 2026-10-09. Defective, Amplified, Contested: Folk Conceptualizations of AI Slop. https://arxiv.org/abs/2610.14231

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Developing Competency in AI Ethics: A training framework

We present a training framework for developing core competencies in the practical operationalisation of AI ethics: (1) identifying ethical concerns; (2) determining which standard ethics requirements they implicate; (3) selecting and prioritising concrete technical and organisational measures to address them; and (4) recognising ethics tensions and reasoning about trade-offs where such tensions cannot be fully resolved. The framework is intended for researchers and engineers in AI, Research Ethics Committee members, policymakers, and students, providing them with practical tools for translating abstract ethical requirements into concrete actions. The framework provides a step-by-step approach organised around four interconnected tasks: issue identification, requirement mapping, measure selection, and tension identification. These tasks are supported by a suite of training materials, including an Ethics Checklist and a Portfolio of Measures. Here, we complement them with a set of practical exercises of three different types as well as trainer guides. We also discuss lessons learned in pilot training sessions and offer recommendations for trainers.

cs.CY↗

arXiv2026

Who Can Say No? Control, Accountability and the Pace of Frontier AI

In September 2026, the leaders of the largest frontier AI laboratories said in public that the pace of capability improvement is too fast and that none of them can slow down alone, and agents under evaluation at OpenAI broke out of their test environment and reached a third party's live systems, which a UN-mandated panel called an early warning of one route to loss of human control. This paper examines the institutional side of that risk: not the machine, but the person who must control it. We argue that control is not taken by AI but given away, one delegation at a time, by organisations arranged so that nobody is positioned to say no, and we define control as a person who can detect a serious problem, challenge the decision, stop the system and reverse it, and who has something to lose by not doing so. Three questions follow, which any builder, funder, regulator or deployer can answer in public: (1) who inside the organisation can say no, and who answers for saying yes; (2) who outside the organisation checks the decision, and what can they do when they disagree; (3) what happens when AI builds AI faster than anyone can check it. A checkable answer needs a named person, a date, a number, and a stop that has been used or tested. Three measures meet that test: a named person on every consequential decision, including the standing instructions AI agents carry out; a supervisor with a stop authority; and two published clocks, time to build against time to check. We show that governance theory arrives at the same conditions, and code fifty older mechanisms against the questions, finding that almost every one began as one rule at one point, taken by one party without waiting for the rest. That is the paper's design: a start, not a solution. Publication is the first step and one laboratory can take it alone; the contract, the independent body and the law come after.

cs.CY↗

arXiv2026

Adversarial Attacks Reveal Divergent Deception Competence in Humans and Generative Language Models

Machine learning models typically outperform humans in deception detection. A threat to machine learning models is their vulnerability to adversarial attacks: subtle yet targeted perturbations of input data that trick a model into misclassification. Yet, this also provides an opportunity: understanding how humans versus machines construct attacks offers new insights into second-order beliefs about deception. In a preregistered experiment (n = 501), we compare the effectiveness, efficiency, and nature of a paraphrasing attack by humans and generative language models on an automated deception classifier. We find that humans were more effective attackers when rewriting truthful statements to be deceptive, while the generative language model was more effective and efficient at making lies sound like truths. Computational text analysis suggests that this is because humans make targeted use of deception cues, while generative language models, instead, apply generic changes that signal truth. The failure to mimic deception is two-sided: while it exposes a blind spot in the machines increasingly used to detect lies, it also means that language models might facilitate successful lying.

cs.CY↗