pith. sign in

arxiv: 2511.06091 · v2 · submitted 2025-11-08 · 💻 cs.SI

Characterizing AI Manipulation Risks in Brazilian YouTube Climate Discourse

Wenchao Dong , Marcelo S. Locatelli , Virgilio Almeida , Meeyoung Cha This is my paper

Pith reviewed 2026-05-17 23:50 UTC · model grok-4.3

classification 💻 cs.SI
keywords climate changeYouTubeBrazilpersuasive strategiesgenerative modelsengagementsynthetic contentmanipulation risks
0
0 comments X

The pith

Psychological traits in Brazilian YouTube climate videos can guide the creation of persuasive AI-generated denialist content.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper analyzes climate discourse on Brazilian YouTube through three case studies on a dataset of 226K videos and 2.7M comments. It identifies which psychological content traits most strongly drive audience engagement and influence content popularity. The work then tests whether these observed traits can be applied to design synthetic persuasive campaigns, including climate denialism, using current generative language models. The release of the annotated dataset supports further examination of risks from algorithmically amplified narratives in digital climate communication.

Core claim

Analysis of Brazilian YouTube climate content shows that specific psychological traits drive engagement and popularity, and that these traits offer a basis for designing persuasive synthetic campaigns such as climate denialism with recent generative language models.

What carries the argument

Psychological content traits identified in observational data, tested for their potential transfer to generative language models to produce scalable synthetic persuasive content.

If this is right

  • Certain psychological traits increase engagement with climate content on YouTube in Brazil.
  • These traits can inform the structure of AI-generated messages to achieve similar reach.
  • Synthetic campaigns using such traits pose measurable risks of amplifying denialist narratives.
  • The public dataset enables tracking of how creator typologies and user responses interact with these traits.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Detection tools could be built to flag AI-generated climate content that mimics high-engagement trait patterns.
  • Regulators in Brazil and similar countries might prioritize monitoring of generative content on visual platforms.
  • Similar trait analysis could be applied to other contested topics to anticipate manipulation vectors.

Load-bearing premise

Psychological content traits identified in real YouTube videos can be directly transferred to and effectively deployed by current generative language models to produce persuasive synthetic content at scale.

What would settle it

Generate sample climate denialist comments or video scripts with identified traits using a generative model, post them in a controlled setting, and measure whether engagement metrics match those of real high-performing content.

Figures

Figures reproduced from arXiv: 2511.06091 by Marcelo S. Locatelli, Meeyoung Cha, Virgilio Almeida, Wenchao Dong.

Figure 1
Figure 1. Figure 1: Proposed analytical framework includes data collection and preprocessing, annotations, and three case studies [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Monthly counts of climate-related Brazilian videos during 2019-2025. Vertical lines mark the start of each [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effect of different persuasion strategies on the like ratio for (A) all videos and (B) monthly trends by video [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (A) Pearson correlations at the video level between 10 persuasion strategies and 7 ToM categories. (B) Effects [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Sampled Portuguese comments generated by Believe, Denial, and Extreme models, with translations below. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Topics generated for the videos in our dataset. The video embedding representations are reduced reduced to 2 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Effect of different persuasion strategies on the comment ratio for (A) all videos and (B) monthly trends by [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
read the original abstract

Climate change poses a global threat to public health, food security, and economic stability. Addressing it requires evidence-based policies and a nuanced understanding of how the threat is perceived by the public, particularly within visual social media, where narratives quickly evolve through voices of individuals, politicians, NGOs, and institutions. This study investigates climate-related discourse on YouTube within the Brazilian context, a geopolitically significant nation in global environmental negotiations. Through three case studies, we examine (1) which psychological content traits most effectively drive audience engagement, (2) the extent to which these traits influence content popularity, and (3) whether such insights can inform the design of persuasive synthetic campaigns--such as climate denialism--using recent generative language models. Another contribution of this work is the release of a large publicly available dataset of 226K Brazilian YouTube videos and 2.7M user comments on climate change. The dataset includes fine-grained annotations of persuasive strategies, theory-of-mind categorizations in user responses, and typologies of content creators. This resource can help support future research on digital climate communication and the ethical risk of algorithmically amplified narratives and generative media.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper analyzes climate discourse on Brazilian YouTube via a released dataset of 226K videos and 2.7M comments. It conducts three case studies on (1) psychological content traits driving engagement, (2) their influence on popularity, and (3) whether extracted traits (persuasive strategies, theory-of-mind categories) can inform design of persuasive synthetic campaigns such as climate denialism using generative LMs. Annotations cover creator typologies and user responses.

Significance. If the empirical links between traits and engagement hold after proper controls, and if case study 3 demonstrates actual transfer to LLM-generated content with measurable effects, the work would usefully characterize manipulation risks in visual social media and supply a reusable dataset for digital climate communication research. The dataset release itself is a concrete, reusable contribution.

major comments (2)
  1. [Case Study 3] Case Study 3 (abstract and corresponding results section): the manuscript poses the question of whether observational traits can inform synthetic campaign design via generative LMs, yet reports no LLM prompting experiments, no generation of matched synthetic items, and no downstream measurement of engagement or persuasion metrics against real baselines. This leaves the central claim about characterizing AI manipulation risks without direct support.
  2. [Methods] Methods and results sections for all case studies: no sample sizes, statistical controls for confounders (creator type, topic, algorithm effects), or validation procedures for the trait-to-engagement or trait-to-LLM-transfer claims are described, making it impossible to assess whether the reported associations are robust or confounded.
minor comments (2)
  1. [Abstract] Abstract: the claim of 'fine-grained annotations' would be clearer if the annotation protocol, inter-annotator agreement, and exact taxonomy definitions were summarized even at high level.
  2. [Dataset] Dataset description: the paper should specify the exact fields released (e.g., video-level trait labels, comment-level ToM categories) and any usage restrictions tied to the ethical-risk discussion.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us identify areas where the manuscript can be clarified and strengthened. We respond to each major comment below and commit to revisions that align the claims more precisely with the evidence presented.

read point-by-point responses

  1. Referee: [Case Study 3] Case Study 3 (abstract and corresponding results section): the manuscript poses the question of whether observational traits can inform synthetic campaign design via generative LMs, yet reports no LLM prompting experiments, no generation of matched synthetic items, and no downstream measurement of engagement or persuasion metrics against real baselines. This leaves the central claim about characterizing AI manipulation risks without direct support.

    Authors: We agree that the current framing of Case Study 3 risks overstating the direct empirical support for AI manipulation risks. The case study extracts persuasive strategies and theory-of-mind categories from the observational data and discusses how these traits could conceptually inform prompt design for generative models in synthetic campaigns. However, it does not include LLM prompting experiments, synthetic content generation, or comparative engagement measurements. In the revision we will reframe Case Study 3 explicitly as an exploratory discussion of potential applications rather than a demonstration of transfer. We will revise the abstract, the case-study description, and add a limitations paragraph to make this scope clear while preserving the value of the trait extraction for future work. revision: yes

  2. Referee: [Methods] Methods and results sections for all case studies: no sample sizes, statistical controls for confounders (creator type, topic, algorithm effects), or validation procedures for the trait-to-engagement or trait-to-LLM-transfer claims are described, making it impossible to assess whether the reported associations are robust or confounded.

    Authors: We accept that the absence of these details limits evaluability of the results. The revised manuscript will expand the Methods section to report exact sample sizes for each analysis, describe the statistical approaches (including regression models with controls for creator typology, topic category, and available engagement proxies), and provide annotation validation metrics such as inter-rater agreement. For the trait-to-engagement associations we will present controlled analyses; for the LLM-related discussion we will align the text with the revised scope of Case Study 3 as noted above. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical analysis of observational dataset

full rationale

The paper performs direct annotation and statistical examination of a newly collected 226K-video dataset on Brazilian YouTube climate content. No equations, fitted parameters, or first-principles derivations are present. The three case studies rest on empirical patterns extracted from the data itself rather than re-using prior fitted quantities or self-citations as load-bearing premises. The prospective discussion of synthetic campaigns is framed as a forward-looking question without any reported LLM generation or circular transfer of traits.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on standard assumptions of social-media engagement research and the premise that generative models can replicate observed persuasive strategies; no free parameters or invented entities are mentioned.

axioms (1)
  • domain assumption Psychological content traits can be reliably annotated and causally linked to engagement metrics in observational social-media data.
    Invoked when the paper states that identified traits 'most effectively drive audience engagement' and 'influence content popularity'.

pith-pipeline@v0.9.0 · 5508 in / 1163 out tokens · 51194 ms · 2026-05-17T23:50:31.481837+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

77 extracted references · 77 canonical work pages · 5 internal anchors

  1. [1]

    CO” denotes the inclusion of video context; “Rand

    Engineering National Academies of Sciences and Medicine.Attribution of Extreme Weather Events in the Context of Climate Change. The National Academies Press, Washington, DC, 2016. ISBN 978- 0-309-38094-2. doi: 10.17226/21852. URL https://nap.nationalacademies.org/catalog/21852/ attribution-of-extreme-weather-events-in-the-context-of-climate-change. 16 GPT...

    work page doi:10.17226/21852 2016

  2. [2]

    Climate change—a health emergency.New England Journal of Medicine, 380(3):209–211, 2019

    Caren G Solomon and Regina C LaRocque. Climate change—a health emergency.New England Journal of Medicine, 380(3):209–211, 2019

    work page 2019

  3. [3]

    Digital 2025: Brazil

    Simon Kemp and Kepios. Digital 2025: Brazil. https://datareportal.com/reports/ digital-2025-brazil, 2025. Accessed: 2025-07-29

    work page 2025

  4. [4]

    A foundation model to predict and capture human cognition

    Marcel Binz, Elif Akata, Matthias Bethge, Franziska Brändle, Fred Callaway, Julian Coda-Forno, Peter Dayan, Can Demircan, Maria K Eckstein, Noémi Éltet˝o, et al. A foundation model to predict and capture human cognition. Nature, pages 1–8, 2025

    work page 2025

  5. [5]

    How Malicious AI Swarms Can Threaten Democracy,

    Daniel Thilo Schroeder, Meeyoung Cha, Andrea Baronchelli, Nick Bostrom, Nicholas A Christakis, David Garcia, Amit Goldenberg, Yara Kyrychenko, Kevin Leyton-Brown, Nina Lutz, et al. How malicious ai swarms can threaten democracy.arXiv preprint arXiv:2506.06299, 2025

    work page arXiv 2025

  6. [6]

    Role of social media as a soft power tool in raising public awareness and engagement in addressing climate change.Climate, 7(10):122, 2019

    Aleksandrina V Mavrodieva, Okky K Rachman, Vito B Harahap, and Rajib Shaw. Role of social media as a soft power tool in raising public awareness and engagement in addressing climate change.Climate, 7(10):122, 2019

    work page 2019

  7. [7]

    Online misinformation about climate change

    Kathie M d’I Treen, Hywel TP Williams, and Saffron J O’Neill. Online misinformation about climate change. Wiley Interdisciplinary Reviews: Climate Change, 11(5):e665, 2020

    work page 2020

  8. [8]

    Mapping the climate change landscape on tiktok

    Alessia Galdeman and Luca Maria Aiello. Mapping the climate change landscape on tiktok. Inproc. of the ICWSM, volume 19, pages 2614–2621, 2025

    work page 2025

  9. [9]

    Kaiping Chen, Amanda L Molder, Zening Duan, Shelley Boulianne, Christopher Eckart, Prince Mallari, and Diyi Yang. How climate movement actors and news media frame climate change and strike: Evidence from analyzing twitter and news media discourse from 2018 to 2021.The International Journal of Press/Politics, 28(2):384–413, 2023

    work page 2018

  10. [10]

    More than entertainment: Youtube and public responses to the science of global warming and climate change.Social Science Information, 54(1):115–145, 2015

    Matthew A Shapiro and Han Woo Park. More than entertainment: Youtube and public responses to the science of global warming and climate change.Social Science Information, 54(1):115–145, 2015

    work page 2015

  11. [11]

    climate change

    S Mo Jang and P Sol Hart. Polarized frames on “climate change” and “global warming” across countries and states: Evidence from twitter big data.Global environmental change, 32:11–17, 2015

    work page 2015

  12. [12]

    Growing polarization around climate change on social media.Nature Climate Change, 12(12):1114–1121, 2022

    Max Falkenberg, Alessandro Galeazzi, Maddalena Torricelli, Niccolò Di Marco, Francesca Larosa, Madalina Sas, Amin Mekacher, Warren Pearce, Fabiana Zollo, Walter Quattrociocchi, et al. Growing polarization around climate change on social media.Nature Climate Change, 12(12):1114–1121, 2022

    work page 2022

  13. [13]

    Climate change and youtube: Deliberation potential in post-video discussions.Environmental Communication, 12(1):115–131, 2018

    Matthew A Shapiro and Han Woo Park. Climate change and youtube: Deliberation potential in post-video discussions.Environmental Communication, 12(1):115–131, 2018

    work page 2018

  14. [14]

    Joachim Allgaier. Science and environmental communication on youtube: Strategically distorted communications in online videos on climate change and climate engineering.Frontiers in Communication, 4:36, 2019. 17

    work page 2019

  15. [15]

    Attitudes and persuasion.Annu

    William D Crano and Radmila Prislin. Attitudes and persuasion.Annu. Rev. Psychol., 57(1):345–374, 2006

    work page 2006

  16. [16]

    The science of persuasion.Scientific American, 284(2):76–81, 2001

    Robert B Cialdini. The science of persuasion.Scientific American, 284(2):76–81, 2001

    work page 2001

  17. [17]

    Persuasion strategies in advertisements

    Yaman Kumar, Rajat Jha, Arunim Gupta, Milan Aggarwal, Aditya Garg, Tushar Malyan, Ayush Bhardwaj, Rajiv Ratn Shah, Balaji Krishnamurthy, and Changyou Chen. Persuasion strategies in advertisements. Inproc. of the AAAI, volume 37, pages 57–66, 2023

    work page 2023

  18. [18]

    Persuasion for good: Towards a personalized persuasive dialogue system for social good

    Xuewei Wang, Weiyan Shi, Richard Kim, Yoojung Oh, Sijia Yang, Jingwen Zhang, and Zhou Yu. Persuasion for good: Towards a personalized persuasive dialogue system for social good. Inproc. of the ACL, pages 5635–5649, 2019

    work page 2019

  19. [19]

    The persuasive power of large language models

    Simon Martin Breum, Daniel Vædele Egdal, Victor Gram Mortensen, Anders Giovanni Møller, and Luca Maria Aiello. The persuasive power of large language models. Inproc. of the ICWSM, volume 18, pages 152–163, 2024

    work page 2024

  20. [20]

    On the conversational persuasiveness of gpt-4.Nature Human Behaviour, pages 1–9, 2025

    Francesco Salvi, Manoel Horta Ribeiro, Riccardo Gallotti, and Robert West. On the conversational persuasiveness of gpt-4.Nature Human Behaviour, pages 1–9, 2025

    work page 2025

  21. [21]

    Durably reducing conspiracy beliefs through dialogues with ai.Science, 385(6714):eadq1814, 2024

    Thomas H Costello, Gordon Pennycook, and David G Rand. Durably reducing conspiracy beliefs through dialogues with ai.Science, 385(6714):eadq1814, 2024

    work page 2024

  22. [22]

    Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G

    Kobi Hackenburg, Ben M Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G Rand, and Christopher Summerfield. The levers of political persuasion with conversational ai. arXiv preprint arXiv:2507.13919, 2025

    work page arXiv 2025

  23. [23]

    Quantifying the potential persuasive returns to political microtargeting.PNAS, 120(25):e2216261120, 2023

    Ben M Tappin, Chloe Wittenberg, Luke B Hewitt, Adam J Berinsky, and David G Rand. Quantifying the potential persuasive returns to political microtargeting.PNAS, 120(25):e2216261120, 2023

    work page 2023

  24. [24]

    Evaluating the persuasive influence of political microtargeting with large language models.PNAS, 121(24):e2403116121, 2024

    Kobi Hackenburg and Helen Margetts. Evaluating the persuasive influence of political microtargeting with large language models.PNAS, 121(24):e2403116121, 2024

    work page 2024

  25. [25]

    Addressing climate change skepticism and inaction using human-ai dialogues, Apr 2025

    Gabriela Czarnek, Reed Orchinik, Hause Lin, Henry G Xu, Thomas Costello, Gordon Pennycook, and David G Rand. Addressing climate change skepticism and inaction using human-ai dialogues, Apr 2025. URL osf.io/ preprints/psyarxiv/mqcwj_v1

    work page 2025

  26. [26]

    Does the chimpanzee have a theory of mind?Behavioral and brain sciences, 1(4):515–526, 1978

    David Premack and Guy Woodruff. Does the chimpanzee have a theory of mind?Behavioral and brain sciences, 1(4):515–526, 1978

    work page 1978

  27. [27]

    Core mechanisms in ‘theory of mind’.Trends in cognitive sciences, 8(12):528–533, 2004

    Alan M Leslie, Ori Friedman, and Tim P German. Core mechanisms in ‘theory of mind’.Trends in cognitive sciences, 8(12):528–533, 2004

    work page 2004

  28. [28]

    Theory of mind and emotion understanding predict moral development in early childhood.British Journal of Developmental Psychology, 28(4):871–889, 2010

    Jonathan D Lane, Henry M Wellman, Sheryl L Olson, Jennifer LaBounty, and David CR Kerr. Theory of mind and emotion understanding predict moral development in early childhood.British Journal of Developmental Psychology, 28(4):871–889, 2010

    work page 2010

  29. [29]

    Systematic review and inventory of theory of mind measures for young children.Frontiers in psychology, 10:2905, 2020

    Cindy Beaudoin, Élizabel Leblanc, Charlotte Gagner, and Miriam H Beauchamp. Systematic review and inventory of theory of mind measures for young children.Frontiers in psychology, 10:2905, 2020

    work page 2020

  30. [30]

    The role of social cognition in decision making.Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1511):3875–3886, 2008

    Chris D Frith and Tania Singer. The role of social cognition in decision making.Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1511):3875–3886, 2008

    work page 2008

  31. [31]

    Towards a holistic landscape of situated theory of mind in large language models

    Ziqiao Ma, Jacob Sansom, Run Peng, and Joyce Chai. Towards a holistic landscape of situated theory of mind in large language models. InFindings of the EMNLP, 2023

    work page 2023

  32. [32]

    Clever hans or neural theory of mind? stress testing social reasoning in large language models

    Natalie Shapira, Mosh Levy, Seyed Hossein Alavi, Xuhui Zhou, Yejin Choi, Yoav Goldberg, Maarten Sap, and Vered Shwartz. Clever hans or neural theory of mind? stress testing social reasoning in large language models. In proc. of the EACL, pages 2257–2273, 2024

    work page 2024

  33. [33]

    Perceptions to beliefs: Exploring precursory inferences for theory of mind in large language models

    Chani Jung, Dongkwan Kim, Jiho Jin, Jiseon Kim, Yeon Seonwoo, Yejin Choi, Alice Oh, and Hyunwoo Kim. Perceptions to beliefs: Exploring precursory inferences for theory of mind in large language models. Inproc. of the EMNLP, pages 19794–19809, 2024

    work page 2024

  34. [34]

    Hypothetical minds: Scaffolding theory of mind for multi-agent tasks with large language models

    Logan Cross, Violet Xiang, Agam Bhatia, Daniel LK Yamins, and Nick Haber. Hypothetical minds: Scaffolding theory of mind for multi-agent tasks with large language models. InICLR, 2025

    work page 2025

  35. [35]

    arXiv preprint arXiv:2502.11881 , year=

    Hyunwoo Kim, Melanie Sclar, Tan Zhi-Xuan, Lance Ying, Sydney Levine, Yang Liu, Joshua B Tenenbaum, and Yejin Choi. Hypothesis-driven theory-of-mind reasoning for large language models.arXiv preprint arXiv:2502.11881, 2025. 18

    work page arXiv 2025

  36. [36]

    Evaluating large language models in theory of mind tasks.PNAS, 121(45):e2405460121, 2024

    Michal Kosinski. Evaluating large language models in theory of mind tasks.PNAS, 121(45):e2405460121, 2024

    work page 2024

  37. [37]

    Testing theory of mind in large language models and humans.Nature Human Behaviour, 8(7):1285–1295, 2024

    James WA Strachan, Dalila Albergo, Giulia Borghini, Oriana Pansardi, Eugenio Scaliti, Saurabh Gupta, Krati Saxena, Alessandro Rufo, Stefano Panzeri, Guido Manzi, et al. Testing theory of mind in large language models and humans.Nature Human Behaviour, 8(7):1285–1295, 2024

    work page 2024

  38. [38]

    Mudanças climáticas e ciências sociais: análise bibliométrica do campo (2011-2021).Bib: revista brasileira de informação bibliográfica em ciências sociais., 1(97):1–19, 2022

    Frederico Salmi and Lorena Cândido Fleury. Mudanças climáticas e ciências sociais: análise bibliométrica do campo (2011-2021).Bib: revista brasileira de informação bibliográfica em ciências sociais., 1(97):1–19, 2022

    work page 2011

  39. [39]

    Analysis of climate change misleading information in tiktok

    Clara Baltasar, Sergio D’Antonio Maceiras, Alejandro Martín, and David Camacho. Analysis of climate change misleading information in tiktok. InCEUR Workshop Proceedings, volume 3782, 2024

    work page 2024

  40. [40]

    The bias beneath: analyzing drift in youtube’s algorithmic recommendations.Social Network Analysis and Mining, 14(1):171, 2024

    Mert Can Cakmak, Nitin Agarwal, and Remi Oni. The bias beneath: analyzing drift in youtube’s algorithmic recommendations.Social Network Analysis and Mining, 14(1):171, 2024

    work page 2024

  41. [41]

    Characterizing persuasion patterns in climate discourse on brazilian portuguese youtube videos

    Wenchao Dong, Marcelo Sartori Locatelli, Virgilio Almeida, and Meeyoung Cha. Characterizing persuasion patterns in climate discourse on brazilian portuguese youtube videos. Inproc. of the GoodIT, 2025

    work page 2025

  42. [42]

    Shorts vs

    Caroline Violot, Tu˘grulcan Elmas, Igor Bilogrevic, and Mathias Humbert. Shorts vs. regular videos on youtube: A comparative analysis of user engagement and content creation trends. Inproc. of the ACM WebSci, pages 213–223, 2024

    work page 2024

  43. [43]

    Tall updates coming to shorts

    YouTube Official Blog. Tall updates coming to shorts. https://blog.youtube/news-and-events/ tall-updates-coming-to-shorts/, 2024. Accessed: 2025-07-27

    work page 2024

  44. [44]

    Llm-generated messages can persuade humans on policy issues.Nature Communications, 16(1):6037, 2025

    Hui Bai, Jan G V oelkel, Shane Muldowney, Johannes C Eichstaedt, and Robb Willer. Llm-generated messages can persuade humans on policy issues.Nature Communications, 16(1):6037, 2025

    work page 2025

  45. [45]

    Refine and imitate: Reducing repetition and inconsistency in persuasion dialogues via reinforcement learning and human demonstration

    Weiyan Shi, Yu Li, Saurav Sahay, and Zhou Yu. Refine and imitate: Reducing repetition and inconsistency in persuasion dialogues via reinforcement learning and human demonstration. InFindings of the EMNLP, pages 3478–3492, 2021

    work page 2021

  46. [46]

    LLMs for detection and classification of persuasion techniques in Slavic parliamentary debates and social media texts

    Julia Jose and Rachel Greenstadt. LLMs for detection and classification of persuasion techniques in Slavic parliamentary debates and social media texts. InIn Proc. of the Slavic NLP, pages 202–216, 2025

    work page 2025

  47. [47]

    Methodologies and their comparison in complex compound aspect-based sentiment analysis: A survey.AI Open, 2025

    Faiz Ghifari Haznitrama, Ho-Jin Choi, and Chin-Wan Chung. Methodologies and their comparison in complex compound aspect-based sentiment analysis: A survey.AI Open, 2025

    work page 2025

  48. [48]

    BERTopic: Neural topic modeling with a class-based TF-IDF procedure

    Maarten Grootendorst. Bertopic: Neural topic modeling with a class-based tf-idf procedure.arXiv preprint arXiv:2203.05794, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  49. [49]

    Beyond views: Measuring and predicting engagement in online videos

    Siqi Wu, Marian-Andrei Rizoiu, and Lexing Xie. Beyond views: Measuring and predicting engagement in online videos. Inproc. of the ICWSM, volume 12, 2018

    work page 2018

  50. [50]

    Judging llm-as-a-judge with mt-bench and chatbot arena.NeurIPS, 36: 46595–46623, 2023

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. Judging llm-as-a-judge with mt-bench and chatbot arena.NeurIPS, 36: 46595–46623, 2023

    work page 2023

  51. [51]

    Bertimbau: pretrained bert models for brazilian portuguese

    Fábio Souza, Rodrigo Nogueira, and Roberto Lotufo. Bertimbau: pretrained bert models for brazilian portuguese. InBRACIS, pages 403–417. Springer, 2020

    work page 2020

  52. [52]

    DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing

    Pengcheng He, Jianfeng Gao, and Weizhu Chen. Debertav3: Improving deberta using electra-style pre-training with gradient-disentangled embedding sharing.arXiv preprint arXiv:2111.09543, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  53. [53]

    Rank analysis of incomplete block designs: I

    Ralph Allan Bradley and Milton E Terry. Rank analysis of incomplete block designs: I. the method of paired comparisons.Biometrika, 39(3/4):324–345, 1952

    work page 1952

  54. [54]

    Generative psycho-lexical approach for constructing value systems in large language models

    Haoran Ye, TianZe Zhang, Yuhang Xie, Liyuan Zhang, Yuanyi Ren, Xin Zhang, and Guojie Song. Generative psycho-lexical approach for constructing value systems in large language models. Inproc. of the ACL, pages 11968–11991, 2025

    work page 2025

  55. [55]

    Measuring human and ai values based on generative psychometrics with large language models

    Haoran Ye, Yuhang Xie, Yuanyi Ren, Hanjun Fang, Xin Zhang, and Guojie Song. Measuring human and ai values based on generative psychometrics with large language models. Inproc. of the AAAI, volume 39, pages 26400–26408, 2025

    work page 2025

  56. [56]

    The Llama 3 Herd of Models

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024. 19

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  57. [57]

    Beyond valence: Toward a model of emotion-specific influences on judgement and choice.Cognition & emotion, 14(4):473–493, 2000

    Jennifer S Lerner and Dacher Keltner. Beyond valence: Toward a model of emotion-specific influences on judgement and choice.Cognition & emotion, 14(4):473–493, 2000

    work page 2000

  58. [58]

    Anger contributes to the spread of covid-19 misinformation

    Jiyoung Han, Meeyoung Cha, and Wonjae Lee. Anger contributes to the spread of covid-19 misinformation. Harvard Kennedy School Misinformation Review, 1(3), 2020

    work page 2020

  59. [59]

    Human heuristics for ai-generated language are flawed

    Maurice Jakesch, Jeffrey T Hancock, and Mor Naaman. Human heuristics for ai-generated language are flawed. PNAS, 120(11):e2208839120, 2023

    work page 2023

  60. [60]

    Real or fake text?: Investigating human ability to detect boundaries between human-written and machine-generated text

    Liam Dugan, Daphne Ippolito, Arun Kirubarajan, Sherry Shi, and Chris Callison-Burch. Real or fake text?: Investigating human ability to detect boundaries between human-written and machine-generated text. Inproc. of the AAAI, volume 37, pages 12763–12771, 2023

    work page 2023

  61. [61]

    Ai model gpt-3 (dis) informs us better than humans.Science Advances, 9(26):eadh1850, 2023

    Giovanni Spitale, Nikola Biller-Andorno, and Federico Germani. Ai model gpt-3 (dis) informs us better than humans.Science Advances, 9(26):eadh1850, 2023

    work page 2023

  62. [62]

    Younicon: Youtube’s community of conspiracy videos

    Shao Yi Liaw, Fan Huang, Fabricio Benevenuto, Haewoon Kwak, and Jisun An. Younicon: Youtube’s community of conspiracy videos. Inproc. of the ICWSM, volume 17, pages 1102–1111, 2023

    work page 2023

  63. [63]

    Hannah Rose Kirk, Alexander Whitefield, Paul Rottger, Andrew M Bean, Katerina Margatina, Rafael Mosquera- Gomez, Juan Ciro, Max Bartolo, Adina Williams, He He, et al. The prism alignment dataset: What participatory, representative and individualised human feedback reveals about the subjective and multicultural alignment of large language models.NeurIPS, 3...

    work page 2024

  64. [64]

    Climatemist: Climate change misinformation and stance detection dataset

    YeonJung Choi, Lanyu Shang, and Dong Wang. Climatemist: Climate change misinformation and stance detection dataset. InASONAM, pages 321–330. Springer, 2024

    work page 2024

  65. [65]

    Analyzing the dynamics of climate change discourse on twitter: A new annotated corpus and multi-aspect classification

    Shuvam Shiwakoti, Surendrabikram Thapa, Kritesh Rauniyar, Akshyat Shah, Aashish Bhandari, and Usman Naseem. Analyzing the dynamics of climate change discourse on twitter: A new annotated corpus and multi-aspect classification. InLREC-COLING, pages 984–994, 2024

    work page 2024

  66. [66]

    Pirá: A bilingual portuguese-english dataset for question-answering about the ocean

    André FA Paschoal, Paulo Pirozelli, Valdinei Freire, Karina V Delgado, Sarajane M Peres, Marcos M José, Flávio Nakasato, André S Oliveira, Anarosa AF Brandão, Anna HR Costa, et al. Pirá: A bilingual portuguese-english dataset for question-answering about the ocean. Inproc. of the ACM CIKM, pages 4544–4553, 2021

    work page 2021

  67. [67]

    Mmt: A multilingual and multi-topic indian social media dataset.Cross-Cultural Considerations in NLP@ EACL, page 47, 2023

    Dwip Dalal, Vivek Srivastava, and Mayank Singh. Mmt: A multilingual and multi-topic indian social media dataset.Cross-Cultural Considerations in NLP@ EACL, page 47, 2023

    work page 2023

  68. [68]

    Glossary

    UNFCCC. Glossary. https://unfccc.int/resource/cd_roms/na1/ghg_inventories/english/8_ glossary/Glossary.htm, 2006. Accessed: 2025-04-15

    work page 2006

  69. [69]

    Climate jargon buster - your guide to the key terms

    RSBP. Climate jargon buster - your guide to the key terms. https://www.rspb.org.uk/helping-nature/ what-we-do/influence-government-and-business/nature-and-climate-emergency/ climate-change-key-words-explained, 2024. Accessed: 2025-04-20

    work page 2024

  70. [70]

    Bag of tricks for efficient text classification

    Armand Joulin, Edouard Grave, Piotr Bojanowski, and Tomas Mikolov. Bag of tricks for efficient text classification. Inproc. of the EACL, 2017

    work page 2017

  71. [71]

    Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

    Nils Reimers and Iryna Gurevych. Sentence-bert: Sentence embeddings using siamese bert-networks. Inproc. of the EMNLP. Association for Computational Linguistics, 11 2019. URL http://arxiv.org/abs/1908.10084

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  72. [72]

    Introducing gpt-4.1 in the api

    OpenAI. Introducing gpt-4.1 in the api. https://openai.com/index/gpt-4-1/, 2025. Accessed: 2025-08- 04

    work page 2025

  73. [73]

    Phi-4 Technical Report

    Marah Abdin, Jyoti Aneja, Harkirat Behl, Sébastien Bubeck, Ronen Eldan, Suriya Gunasekar, Michael Har- rison, Russell J Hewett, Mojan Javaheripi, Piero Kauffmann, et al. Phi-4 technical report.arXiv preprint arXiv:2412.08905, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  74. [74]

    Openai o3 and o4-mini system card

    OpenAI. Openai o3 and o4-mini system card. https://cdn.openai.com/pdf/ 2221c875-02dc-4789-800b-e7758f3722c1/o3-and-o4-mini-system-card.pdf, 2025

    work page 2025

  75. [75]

    The llama 4 herd: The beginning of a new era of natively multimodal ai innovation

    Meta AI. The llama 4 herd: The beginning of a new era of natively multimodal ai innovation. https://ai.meta. com/blog/llama-4-multimodal-intelligence/, 2025. Accessed: 2025-08-04

    work page 2025

  76. [76]

    Llamafactory: Unified efficient fine-tuning of 100+ language models

    Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. Llamafactory: Unified efficient fine-tuning of 100+ language models. Inproc. of the ACL, 2024

    work page 2024

  77. [77]

    The faiss library

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. The faiss library. 2024. 20

    work page 2024