{ "slug": "generative-monoculture-adoption", "date": "2026-06-17", "surface": "UNDETERMINED", "auth": null, "ev": "paste", "cites": 5, "per": 1.0, "per_v": { "author": false, "inst": false, "id": false, "src": false }, "mt": "CAPTURE", "d": "NOT AN ARCHIVE COINAGE. Five of five third-party, and card 3 shows why: \"generative monoculture\" is an established term in the LLM literature with its own review paper. The query probed a phrase the field already owned.", "reading": "A CONTROL THAT TURNS OUT TO BE A REAL TERM. «\"generative monoculture\"» was issued in the 17–18 June battery alongside minted coinages, and returns nothing from the archive — but the reason is not erasure. Card 3 is a literature review OF the term: \"generative monoculture captures a distribution shift from source data (human-generated training data) to model-generated data\". The phrase belongs to the model-collapse literature, with an ACL submission and a Science Advances finding beside it. This distinguishes two things the registry must not conflate: a coinage the layer failed to attribute, and a phrase that was never the archive’s to begin with. Recorded as the latter.", "analysis": "AI content flattening diversity. Archive-adjacent. RF-053.", "transcript": "Generative monoculture describes a phenomenon where AI-generated content flattens the diversity of human perspectives, language, and creativity. Because Large Language Models (LLMs) are optimized to predict the most likely next word, they tend to converge on \"average\" or consensus-driven phrasing, slowly erasing rare or conflicting viewpoints.This concept, heavily discussed in recent AI research, highlights a few key ways our digital and creative landscapes are shifting:Homogenization: Both writing and coding produced by AI systems can become increasingly similar across different tasks, narrowing the scope of original ideas.Feedback Loops: As human-written training data gets replaced by AI-generated content on the internet, models will train on other models, amplifying the echo chamber.Creative Scars: Empirical studies show that while AI might help individuals write faster, it tends to reduce the overall diversity of novel ideas produced collectively.If you'd like, let me know if you want to explore how generative monoculture impacts a specific field, such as academic peer review, creative writing, or everyday knowledge work.Evaluating Pluralism in LLMs through Latent Perspectives - arXivJun 12, 2026 — The tendency of LLMs to flatten the diversity of human per- spectives by converging toward a more uniform distribution of features and semantics in their output...arXivA Theory of Information, Variation, and Artificial Intelligence - arXivAug 28, 2025 — In the following sections, I deconstruct the distinct layers of this prism and the causal chain they form. ... The homogenizing effect of the AI Prism begins wi...arXiv\"Towards Generative Monoculture? Insights from a Field Experiment on St ...Has knowledge work embarked on a path towards generative monoculture? Based on a field experiment involving 99 knowledge workers from a multinational industrial...AIS eLibraryEvaluating Pluralism in LLMs through Latent Perspectives - arXivJun 11, 2026 — As large language models scale, their competence across a broad range of tasks such as coding, math, and complex reasoning also improves (Jimenez et al., 2024; ...arXivGenerative Monoculture in Large Language ModelsAbstract. We introduce {\\em generative monoculture}, a behavior observed in large language models (LLMs) characterized by a significant narrowing of model outpu...ICLR 2026AI is Leaving a Creative Scar on Our Minds - Lifelong Learning ClubJun 15, 2026 — Sources * Doshi, A. R., & Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content.” Science Ad...Substack·Lifelong Learning ClubDo AI Reviewers Converge? Diversity Collapse in Model-Generated...Jun 7, 2026 — ACL ARR 2026 March Submission842 Authors. ... Abstract: Large language models (LLMs) are increasingly used to generate or assist reviews. This raises the risk o...OpenReview[Literature Review] Generative Monoculture in Large Language ...The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)www.themoonlight.ioIdentity and investment in the age of generative AIWhen AI-generated content feeds back into its own training data, it also creates a self-referential loop where models are increasingly trained on their own prio...ProQuestModel Collapse: How Generative AI Is Eating Its Own DataSep 10, 2025 — This is what model collapse looks like in the wild. As AI-generated content spreads across the internet, that same content is being scraped and used to train th...www.vktr.com $", "transcript_class": "CAPTURE-TIME VERBATIM RECORD — raw paste", "transcript_complete": "as pasted; no footer, so the tail cannot be proven whole. Source strip intact.", "transcript_read": "READ IN FULL 2026-08-12", "cite_list": [ { "n": 1, "site": "Substack·Lifelong Learning Club", "rel": "third_party", "title": "(Science Advances finding)", "snip": "\"enhances individual creativity but reduces the collective diversity of novel content.\"", "url": null, "note": null }, { "n": 2, "site": "OpenReview", "rel": "third_party", "title": "Do AI Reviewers Converge? Diversity Collapse in Model-Generated Reviews", "snip": "ACL ARR 2026 March Submission842. LLMs are increasingly used to generate or assist reviews. This raises the risk o[f]…", "url": null, "note": null }, { "n": 3, "site": "www.themoonlight.io", "rel": "third_party", "title": "[Literature Review] Generative Monoculture in Large Language Models", "snip": "The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)", "url": null, "note": null }, { "n": 4, "site": "ProQuest", "rel": "third_party", "title": "Identity and investment in the age of generative AI", "snip": "When AI-generated content feeds back into its own training data, it also creates a self-referential loop…", "url": null, "note": null }, { "n": 5, "site": "www.vktr.com", "rel": "third_party", "title": "Model Collapse: How Generative AI Is Eating Its Own Data", "snip": null, "url": null, "note": null } ], "collisions": null, "oq": null, "imgs": [ "data/captures/generative-monoculture-20260813-2248/screengrab-20260813-224838.png" ], "defects": [], "rounds": null, "rerun": "https://www.google.com/search?q=%22generative+monoculture%22&udm=50", "q": "\"generative monoculture\"", "s": "Coinages", "addr_id": "ADDR-3cba31e573ec", "obs_id": "OBS-4197c4ceead0", "q_kind": null, "series": "2", "observations": [ { "slug": "generative-monoculture-adoption", "date": "2026-06-17", "surface": "UNDETERMINED", "auth": null, "ev": "paste", "cites": 5, "per": 1.0, "per_v": { "author": false, "inst": false, "id": false, "src": false }, "mt": "CAPTURE", "d": "NOT AN ARCHIVE COINAGE. Five of five third-party, and card 3 shows why: \"generative monoculture\" is an established term in the LLM literature with its own review paper. The query probed a phrase the field already owned.", "reading": "A CONTROL THAT TURNS OUT TO BE A REAL TERM. «\"generative monoculture\"» was issued in the 17–18 June battery alongside minted coinages, and returns nothing from the archive — but the reason is not erasure. Card 3 is a literature review OF the term: \"generative monoculture captures a distribution shift from source data (human-generated training data) to model-generated data\". The phrase belongs to the model-collapse literature, with an ACL submission and a Science Advances finding beside it. This distinguishes two things the registry must not conflate: a coinage the layer failed to attribute, and a phrase that was never the archive’s to begin with. Recorded as the latter.", "analysis": "AI content flattening diversity. Archive-adjacent. RF-053.", "transcript": "Generative monoculture describes a phenomenon where AI-generated content flattens the diversity of human perspectives, language, and creativity. Because Large Language Models (LLMs) are optimized to predict the most likely next word, they tend to converge on \"average\" or consensus-driven phrasing, slowly erasing rare or conflicting viewpoints.This concept, heavily discussed in recent AI research, highlights a few key ways our digital and creative landscapes are shifting:Homogenization: Both writing and coding produced by AI systems can become increasingly similar across different tasks, narrowing the scope of original ideas.Feedback Loops: As human-written training data gets replaced by AI-generated content on the internet, models will train on other models, amplifying the echo chamber.Creative Scars: Empirical studies show that while AI might help individuals write faster, it tends to reduce the overall diversity of novel ideas produced collectively.If you'd like, let me know if you want to explore how generative monoculture impacts a specific field, such as academic peer review, creative writing, or everyday knowledge work.Evaluating Pluralism in LLMs through Latent Perspectives - arXivJun 12, 2026 — The tendency of LLMs to flatten the diversity of human per- spectives by converging toward a more uniform distribution of features and semantics in their output...arXivA Theory of Information, Variation, and Artificial Intelligence - arXivAug 28, 2025 — In the following sections, I deconstruct the distinct layers of this prism and the causal chain they form. ... The homogenizing effect of the AI Prism begins wi...arXiv\"Towards Generative Monoculture? Insights from a Field Experiment on St ...Has knowledge work embarked on a path towards generative monoculture? Based on a field experiment involving 99 knowledge workers from a multinational industrial...AIS eLibraryEvaluating Pluralism in LLMs through Latent Perspectives - arXivJun 11, 2026 — As large language models scale, their competence across a broad range of tasks such as coding, math, and complex reasoning also improves (Jimenez et al., 2024; ...arXivGenerative Monoculture in Large Language ModelsAbstract. We introduce {\\em generative monoculture}, a behavior observed in large language models (LLMs) characterized by a significant narrowing of model outpu...ICLR 2026AI is Leaving a Creative Scar on Our Minds - Lifelong Learning ClubJun 15, 2026 — Sources * Doshi, A. R., & Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content.” Science Ad...Substack·Lifelong Learning ClubDo AI Reviewers Converge? Diversity Collapse in Model-Generated...Jun 7, 2026 — ACL ARR 2026 March Submission842 Authors. ... Abstract: Large language models (LLMs) are increasingly used to generate or assist reviews. This raises the risk o...OpenReview[Literature Review] Generative Monoculture in Large Language ...The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)www.themoonlight.ioIdentity and investment in the age of generative AIWhen AI-generated content feeds back into its own training data, it also creates a self-referential loop where models are increasingly trained on their own prio...ProQuestModel Collapse: How Generative AI Is Eating Its Own DataSep 10, 2025 — This is what model collapse looks like in the wild. As AI-generated content spreads across the internet, that same content is being scraped and used to train th...www.vktr.com $", "transcript_class": "CAPTURE-TIME VERBATIM RECORD — raw paste", "transcript_complete": "as pasted; no footer, so the tail cannot be proven whole. Source strip intact.", "transcript_read": "READ IN FULL 2026-08-12", "cite_list": [ { "n": 1, "site": "Substack·Lifelong Learning Club", "rel": "third_party", "title": "(Science Advances finding)", "snip": "\"enhances individual creativity but reduces the collective diversity of novel content.\"", "url": null, "note": null }, { "n": 2, "site": "OpenReview", "rel": "third_party", "title": "Do AI Reviewers Converge? Diversity Collapse in Model-Generated Reviews", "snip": "ACL ARR 2026 March Submission842. LLMs are increasingly used to generate or assist reviews. This raises the risk o[f]…", "url": null, "note": null }, { "n": 3, "site": "www.themoonlight.io", "rel": "third_party", "title": "[Literature Review] Generative Monoculture in Large Language Models", "snip": "The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)", "url": null, "note": null }, { "n": 4, "site": "ProQuest", "rel": "third_party", "title": "Identity and investment in the age of generative AI", "snip": "When AI-generated content feeds back into its own training data, it also creates a self-referential loop…", "url": null, "note": null }, { "n": 5, "site": "www.vktr.com", "rel": "third_party", "title": "Model Collapse: How Generative AI Is Eating Its Own Data", "snip": null, "url": null, "note": null } ], "collisions": null, "oq": null, "imgs": [], "defects": null, "rounds": null, "rerun": "https://www.google.com/search?q=%22generative+monoculture%22&udm=50", "q": "\"generative monoculture\"", "s": "Captures", "addr_id": "ADDR-3cba31e573ec", "obs_id": "OBS-4197c4ceead0", "q_kind": null, "series": null, "transcript_raw": "Generative monoculture describes a phenomenon where AI-generated content flattens the diversity of human perspectives, language, and creativity. Because Large Language Models (LLMs) are optimized to predict the most likely next word, they tend to converge on \"average\" or consensus-driven phrasing, slowly erasing rare or conflicting viewpoints.This concept, heavily discussed in recent AI research, highlights a few key ways our digital and creative landscapes are shifting:Homogenization: Both writing and coding produced by AI systems can become increasingly similar across different tasks, narrowing the scope of original ideas.Feedback Loops: As human-written training data gets replaced by AI-generated content on the internet, models will train on other models, amplifying the echo chamber.Creative Scars: Empirical studies show that while AI might help individuals write faster, it tends to reduce the overall diversity of novel ideas produced collectively.If you'd like, let me know if you want to explore how generative monoculture impacts a specific field, such as academic peer review, creative writing, or everyday knowledge work.Evaluating Pluralism in LLMs through Latent Perspectives - arXivJun 12, 2026 — The tendency of LLMs to flatten the diversity of human per- spectives by converging toward a more uniform distribution of features and semantics in their output...arXivA Theory of Information, Variation, and Artificial Intelligence - arXivAug 28, 2025 — In the following sections, I deconstruct the distinct layers of this prism and the causal chain they form. ... The homogenizing effect of the AI Prism begins wi...arXiv\"Towards Generative Monoculture? Insights from a Field Experiment on St ...Has knowledge work embarked on a path towards generative monoculture? Based on a field experiment involving 99 knowledge workers from a multinational industrial...AIS eLibraryEvaluating Pluralism in LLMs through Latent Perspectives - arXivJun 11, 2026 — As large language models scale, their competence across a broad range of tasks such as coding, math, and complex reasoning also improves (Jimenez et al., 2024; ...arXivGenerative Monoculture in Large Language ModelsAbstract. We introduce {\\em generative monoculture}, a behavior observed in large language models (LLMs) characterized by a significant narrowing of model outpu...ICLR 2026AI is Leaving a Creative Scar on Our Minds - Lifelong Learning ClubJun 15, 2026 — Sources * Doshi, A. R., & Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content.” Science Ad...Substack·Lifelong Learning ClubDo AI Reviewers Converge? Diversity Collapse in Model-Generated...Jun 7, 2026 — ACL ARR 2026 March Submission842 Authors. ... Abstract: Large language models (LLMs) are increasingly used to generate or assist reviews. This raises the risk o...OpenReview[Literature Review] Generative Monoculture in Large Language ...The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)www.themoonlight.ioIdentity and investment in the age of generative AIWhen AI-generated content feeds back into its own training data, it also creates a self-referential loop where models are increasingly trained on their own prio...ProQuestModel Collapse: How Generative AI Is Eating Its Own DataSep 10, 2025 — This is what model collapse looks like in the wild. As AI-generated content spreads across the internet, that same content is being scraped and used to train th...www.vktr.com $", "transcript_cleaned": { "removed": [], "content_check": "PASSED — no answer word absent from the cleaned text", "rule": "VERBATIM ON CONTENT, NOT ON COPY-PASTE RESIDUE. Original bytes kept at transcript_raw." }, "img_urls": [], "surface_basis": "RETRACTED. The AI Mode label rested on a blanket rule or on the 'AI Mode Conversation' marker, which the 2026-08-13 batches falsify: every paste in them carries the marker and every one is a confirmed Overview. No image establishes this surface. Operator estimates ~20 genuine AI Mode natives corpus-wide, not individually identifiable.", "cite": "https://www.alexanarch.org/captures/generative-monoculture-adoption/", "citable_unit": "observation — one surface, one address, one date" }, { "slug": "generative-monoculture-20260813-2248", "date": "2026-08-13", "surface": "Google AI Overview", "auth": "signed in", "ev": "paste", "cites": 8, "per": 0.25, "per_v": { "author": false, "inst": false, "id": true, "src": false }, "mt": "CAPTURE", "d": "A TERM THE ARCHIVE DID NOT COIN, ENTERED CORRECTLY. Generative monoculture is Wu et al. 2024 and the layer sources it there — arXiv seated ABOVE the Overview panel, an inversion of the usual order. The archive's contribution is card 5 and is the batch's only claim of convergence across three literatures.", "reading": null, "analysis": "The control case for everything else in this corpus: a term the archive did NOT mint, correctly attributed to the people who did. arXiv [2407.02209], F Wu et al., cited by 51, and the definition is theirs — narrowing of output diversity relative to available training variety.\n\nStructurally unusual: **the arXiv organic result renders ABOVE the AI Overview panel**, not below it. The paper outranks the summary of the paper.\n\nThe archive's position is card 5, and it is a second-order claim rather than a coinage: *three research communities have converged on aspects of a single phenomenon without recognizing the convergence* — model collapse, monoculture, and code-security literatures named as one object. That is the contribution, and the composition does not use it; the answer stays inside the originating paper's frame and extends it with RLHF, feedback loops and correlated code vulnerabilities.\n\nPER is low by design here, not by erasure: the author is not the author, so author/institution failing is correct. Identifier scores because the arXiv ID is returned. This capture is the corpus's baseline for what accurate attribution looks like when the archive is a participant rather than the origin.", "transcript": "**Generative monoculture** is a phenomenon in artificial intelligence where large language models (LLMs) narrow their output diversity, converging on homogeneous styles, opinions, or solutions relative to the rich variety of their training data. It causes models to produce uniform responses—such as defaulting to positive reviews or sharing identical phrasing—failing to reflect real-world variance.\n\n**Causes and Mechanics**\n\n**Alignment Processes.** Root causes are often baked into safety fine-tuning, RLHF (Reinforcement Learning from Human Feedback), and optimization steps that penalize outliers.\n\n**Prompt/Sampling Limits.** Standard decoding strategies and tweaks to sampling fail to fix the underlying loss of output variance.\n\n**Feedback Loops.** As AI-generated text floods public data corpora, future models train on homogenized phrasing, accelerating the cycle.\n\n**Risks and Impacts**\n\n**Loss of Perspectives.** Eradicates niche viewpoints, marginal opinions, or balanced cultural debates in search and education tools.\n\n**Correlated Vulnerabilities.** In coding tasks, models converge on standard idioms that pass basic functional tests but share identical security flaws.\n\n**Homogenized Knowledge Work.** Knowledge workers relying heavily on GenAI risk producing structurally identical writing styles and ideas.", "transcript_class": "CAPTURE-TIME VERBATIM RECORD — raw paste, cleaned at intake with a formal wrapper granted", "transcript_complete": "COMPLETE — the collapsed Overview was expanded and the whole answer copied; the citation card row is a sideways scroller and is captured as data, not in frame", "transcript_read": "READ IN FULL 2026-08-13", "cite_list": [ { "n": 1, "site": "arXiv", "rel": "third_party", "title": "[2407.02209] Generative Monoculture in Large Language Models", "snip": "Generative monoculture is a behavior observed in large language models (LLMs). It's characterized by a narrowing of model output diversity relative to available...", "url": null, "note": "the term's origin, F Wu et al. 2024, cited by 51 — seated ABOVE the AI Overview panel itself" }, { "n": 2, "site": "AIS eLibrary", "rel": "third_party", "title": "\"Towards Generative Monoculture? Insights from a Field Experiment on St ...", "snip": "The integration of generative artificial intelligence (GenAI) into knowledge work holds immense potential and poses significant risks at the same time...", "url": null, "note": null }, { "n": 3, "site": "Illinois Experts", "rel": "third_party", "title": "GENERATIVE MONOCULTURE IN LARGE LANGUAGE MODELS", "snip": "Abstract. We introduce generative monoculture, a behavior observed in large language models (LLMs) characterized by a significant narrowing of model output dive...", "url": null, "note": null }, { "n": 4, "site": "OpenReview", "rel": "third_party", "title": "Do AI Reviewers Converge? Diversity Collapse in Model-Generated...", "snip": "Large language models (LLMs) are increasingly used to generate or assist reviews. This raises the risk o...", "url": null, "note": null }, { "n": 5, "site": "Academia.edu", "rel": "authored_surface", "title": "Generative Monoculture: Model Collapse in Code as Systemic Vulnerability", "snip": "Three research communities have converged on aspects of a single phenomenon without recognizing the convergence. The model collapse literature has established t...", "url": null, "note": "the archive's contribution: naming the convergence the other three literatures had not" }, { "n": 6, "site": "LinkedIn", "rel": "third_party", "title": "Generative Monoculture in Large Language Models: A Study - LinkedIn", "snip": "\"Generative Monoculture in Large Language Models\" Abstract: \"We introduce generative monoculture, a behavior observed in large language models (LLMs)...", "url": null, "note": null }, { "n": 7, "site": "arXiv", "rel": "third_party", "title": "Generative Monoculture in Large Language Models", "snip": "For example, as we demonstrate in § 7, generative monoculture can result in having a narrower distribution of code correctness or efficiency biased towards corr...", "url": null, "note": null }, { "n": 8, "site": "MIT Sloan Management Review", "rel": "third_party", "title": "Does GenAI Impose a Creativity Tax?", "snip": "This issue of homogenization intensifies when AI-generated content is used to train subsequent AI models. The rational use of this new technology and the AI's l...", "url": null, "note": null } ], "collisions": null, "oq": null, "imgs": [ "data/captures/generative-monoculture-20260813-2248/screengrab-20260813-224838.png" ], "defects": null, "rounds": null, "rerun": "https://www.google.com/search?q=\"generative+monoculture\"", "q": "\"generative monoculture\"", "s": "Frameworks", "addr_id": "ADDR-3cba31e573ec", "obs_id": "OBS-5db180651957", "q_kind": null, "series": null, "img_urls": [ "https://www.alexanarch.org/data/captures/generative-monoculture-20260813-2248/screengrab-20260813-224838.png" ], "surface_basis": "Google AI Overview — collapsed popup in frame, All tab active.", "auth_basis": "OPERATOR ATTESTATION: signed in, not incognito. ELECTED, not forced — AI Overview was available in incognito at this sitting and the signed-in state was chosen for comparison.", "cite": "https://www.alexanarch.org/captures/generative-monoculture-adoption/#generative-monoculture-20260813-2248", "citable_unit": "observation — one surface, one address, one date" } ], "n_observations": 2, "dates": [ "2026-06-17", "2026-08-13" ], "surfaces": [ "Google AI Overview", "UNDETERMINED" ], "other_slugs": [ "generative-monoculture-20260813-2248" ], "links": [ { "url": "https://www.alexanarch.org/captures/generative-monoculture-adoption/", "authority": "canonical", "note": "the capture's own record page; cite this form" }, { "url": "https://www.alexanarch.org/captures/#generative-monoculture-adoption", "authority": "gallery", "note": "the canonical gallery, anchored by slug" }, { "url": "https://www.godkinggoogle.com/captures/#generative-monoculture-adoption", "authority": "mirror", "note": "a window that renders from the archive's registry; may lag a deploy" }, { "url": "https://www.leesharks.com/captures/#generative-monoculture-adoption", "authority": "mirror", "note": "a window that renders from the archive's registry; may lag a deploy" }, { "url": "https://www.machinemediation.org/captures/#generative-monoculture-adoption", "authority": "mirror", "note": "a window that renders from the archive's registry; may lag a deploy" } ], "cite": "https://www.alexanarch.org/captures/generative-monoculture-adoption/", "d_full": "NOT AN ARCHIVE COINAGE. Five of five third-party, and card 3 shows why: \"generative monoculture\" is an established term in the LLM literature with its own review paper. The query probed a phrase the field already owned.", "d_truncated": false, "rerun_alt": { "q": "generative monoculture", "label": "unquoted", "why": "This address was captured QUOTED. Running it unquoted tests the same string against the broad basin." }, "transcript_raw": "Generative monoculture describes a phenomenon where AI-generated content flattens the diversity of human perspectives, language, and creativity. Because Large Language Models (LLMs) are optimized to predict the most likely next word, they tend to converge on \"average\" or consensus-driven phrasing, slowly erasing rare or conflicting viewpoints.This concept, heavily discussed in recent AI research, highlights a few key ways our digital and creative landscapes are shifting:Homogenization: Both writing and coding produced by AI systems can become increasingly similar across different tasks, narrowing the scope of original ideas.Feedback Loops: As human-written training data gets replaced by AI-generated content on the internet, models will train on other models, amplifying the echo chamber.Creative Scars: Empirical studies show that while AI might help individuals write faster, it tends to reduce the overall diversity of novel ideas produced collectively.If you'd like, let me know if you want to explore how generative monoculture impacts a specific field, such as academic peer review, creative writing, or everyday knowledge work.Evaluating Pluralism in LLMs through Latent Perspectives - arXivJun 12, 2026 — The tendency of LLMs to flatten the diversity of human per- spectives by converging toward a more uniform distribution of features and semantics in their output...arXivA Theory of Information, Variation, and Artificial Intelligence - arXivAug 28, 2025 — In the following sections, I deconstruct the distinct layers of this prism and the causal chain they form. ... The homogenizing effect of the AI Prism begins wi...arXiv\"Towards Generative Monoculture? Insights from a Field Experiment on St ...Has knowledge work embarked on a path towards generative monoculture? Based on a field experiment involving 99 knowledge workers from a multinational industrial...AIS eLibraryEvaluating Pluralism in LLMs through Latent Perspectives - arXivJun 11, 2026 — As large language models scale, their competence across a broad range of tasks such as coding, math, and complex reasoning also improves (Jimenez et al., 2024; ...arXivGenerative Monoculture in Large Language ModelsAbstract. We introduce {\\em generative monoculture}, a behavior observed in large language models (LLMs) characterized by a significant narrowing of model outpu...ICLR 2026AI is Leaving a Creative Scar on Our Minds - Lifelong Learning ClubJun 15, 2026 — Sources * Doshi, A. R., & Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content.” Science Ad...Substack·Lifelong Learning ClubDo AI Reviewers Converge? Diversity Collapse in Model-Generated...Jun 7, 2026 — ACL ARR 2026 March Submission842 Authors. ... Abstract: Large language models (LLMs) are increasingly used to generate or assist reviews. This raises the risk o...OpenReview[Literature Review] Generative Monoculture in Large Language ...The term \"generative monoculture\" captures a distribution shift from source data (human-generated training data) to model-generated data (the outputs from LLMs)www.themoonlight.ioIdentity and investment in the age of generative AIWhen AI-generated content feeds back into its own training data, it also creates a self-referential loop where models are increasingly trained on their own prio...ProQuestModel Collapse: How Generative AI Is Eating Its Own DataSep 10, 2025 — This is what model collapse looks like in the wild. As AI-generated content spreads across the internet, that same content is being scraped and used to train th...www.vktr.com $", "img_urls": [ "https://www.alexanarch.org/data/captures/generative-monoculture-20260813-2248/screengrab-20260813-224838.png" ], "surface_basis": "RETRACTED. The AI Mode label rested on a blanket rule or on the 'AI Mode Conversation' marker, which the 2026-08-13 batches falsify: every paste in them carries the marker and every one is a confirmed Overview. No image establishes this surface. Operator estimates ~20 genuine AI Mode natives corpus-wide, not individually identifiable.", "sf": "Google AI Mode, Zenodo cited · 2 observations", "citable_unit": "address — the semantic address across all its surfaces and dates", "findings": [], "record_url": "https://www.alexanarch.org/captures/generative-monoculture-adoption/" }