CoChat
Sign Up
Cited by 61Misinformation and Its ImpactsRecommender Systems and TechniquesSpam and Phishing Detection

Auditing YouTube’s Recommendation Algorithm for Misinformation Filter Bubbles

Open access
Srba, Ivan, Moro, Robert, Tomlein, Matus, Pecher, Branislav, Simko, Jakub, Stefancova, Elena, Kompan, Michal, Hrckova, Andrea +3 more·2023·ACM Transactions on Recommender Systems
DOI 10.1145/3568392
Open this paper in CoChat — free accountOpen PDF

Abstract

In this article, we present results of an auditing study performed over YouTube aimed at investigating how fast a user can get into a misinformation filter bubble, but also what it takes to “burst the bubble,” i.e., revert the bubble enclosure. We employ a sock puppet audit methodology, in which pre-programmed agents (acting as YouTube users) delve into misinformation filter bubbles by watching misinformation-promoting content. Then they try to burst the bubbles and reach more balanced recommendations by watching misinformation-debunking content. We record search results, home page results, and recommendations for the watched videos. Overall, we recorded 17,405 unique videos, out of which we manually annotated 2,914 for the presence of misinformation. The labeled data was used to train a machine learning model classifying videos into three classes (promoting, debunking, neutral) with the accuracy of 0.82. We use the trained model to classify the remaining videos that would not be feasible to annotate manually. Using both the manually and automatically annotated data, we observe the misinformation bubble dynamics for a range of audited topics. Our key finding is that even though filter bubbles do not appear in some situations, when they do, it is possible to burst them by watching misinformation-debunking content (albeit it manifests differently from topic to topic). We also observe a sudden decrease of misinformation filter bubble effect when misinformation-debunking videos are watched after misinformation-promoting videos, suggesting a strong contextuality of recommendations. Finally, when comparing our results with a previous similar study, we do not observe significant improvements in the overall quantity of recommended misinformation content.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1145/3568392

Cites

6 of 6 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Most research fails to adequately address dynamic shifts in algorithmic filtering.

    “Although a few studies have conducted continuous audits to examine how algorithms evolve (e.g., [147,166]), most research fails to address these dynamic shifts adequately.”

    From A systematic review of echo chamber research: comparative analysis of conceptualizations, operationalizations, and varying outcomes (2025) · cites this paper as [147]

    The citation is attached to the sentence’s example of a study conducting continuous audits to examine how algorithms evolve. This paper says instead that it ran its experiments in the shortest possible time frame to minimize the risk of algorithm changes (Methods, p. 12), and describes continuous auditing as future work (Discussion, p. 30). It does identify YouTube’s ongoing development as an underinvestigated issue (Related Work, p. 8), but it is not an example of a continuous audit. (LLM, conf=0.88)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 40

  1. [1][1] Deena Abul-Fottouh, Melodie Yunju Song, and Anatoliy Gruzd. 2020. Examining algorithmic biases in YouTube's recommendations of vaccine videos. Int. Journal of Medical Informatics 140 (2020), 104175. https://doi.org/10.1016/j. ijmedinf.2020.104175 - · DOI
  2. [2][2] Guy Aridor, Duarte Goncalves, and Shan Sikdar. 2020. Deconstructing the Filter Bubble: User Decision-Making and Recommender Systems. In RecSys 2020 - 14th ACM Conference on Recommender Systems . Association for Computing Machinery, Inc, 82-91. https://doi.org/10.1145/3383313.3412246 arXiv:1904.10527 - · DOI
  3. [3][3] Eytan Bakshy, Solomon Messing, and Lada A. Adamic. 2015. Exposure to ideologically diverse news and opinion on Facebook. Science 348, 6239 (2015), 1130-1132. -
  4. [4][4] Cameron Ballard, Ian Goldstein, Pulak Mehta, Genesis Smothers, Kejsi Take, Victoria Zhong, Rachel Greenstadt, Tobias Lauinger, and Damon McCoy. 2022. Conspiracy Brokers: Understanding the Monetization of YouTube Conspiracy Theories. In Proceedings of the ACM Web Conference 2022 (WWW '22) . Association for Computing Machinery, New York, NY, USA, 2707-2718. https://doi.org/10.1145/3485447.3512142 - · DOI
  5. [6][6] Alessandro Bessi. 2016. Personality traits and echo chambers on facebook. Computers in Human Behavior 65 (2016), 319-324. -
  6. [8][8] Laura Burbach, Patrick Halbach, Martina Ziefle, and André Calero Valdez. 2019. Bubble Trouble: Strategies Against Filter Bubbles in Online Social Networks. In Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , Vol. 11582 LNCS. Springer Verlag, 441-456. https: //doi.org/10.1007/978-3-030-22219-2\_33 - · DOI
  7. [9][9] Michał Choraś, Konstantinos Demestichas, Agata Giełczyk, Álvaro Herrero, Paweł Ksieniewicz, Konstantina Remoundou, Daniel Urda, and Michał Woźniak. 2021. Advanced Machine Learning Techniques for Fake News (Online Disinformation) Detection: A Systematic Mapping Study. Applied Soft Computing 101 (March 2021), 107050. https://doi.org/10.1016/j.asoc.2020.107050 arXiv:2101.01142 - · DOI
  8. [10][10] Antrea Chrysanthou, Pinar Barlas, Kyriakos Kyriakou, Styliani Kleanthous, and Jahna Otterbacher. 2020. Bursting the Bubble: Tool for Awareness and Research about Overpersonalization in Information Access Systems. International Conference on Intelligent User Interfaces, Proceedings IUI (2020), 112-113. https://doi.org/10.1145/3379336.3381863 - · DOI
  9. [13][13] European Commission, Content Directorate-General for Communications Networks, and Technology. 2020. The Assessment List for Trustworthy Artificial Intelligence (ALTAI) for self assessment . Publications Office. https://doi.org/ doi/10.2759/791819 - · DOI
  10. [14][14] Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep Neural Networks for YouTube Recommendations. In Proc. of the 10th ACM Conference on Recommender Systems (RecSys '16) . ACM, New York, NY, USA, 191-198. https: //doi.org/10.1145/2959100.2959190 - · DOI · Paper page
  11. [19][19] European Commission. 2018. A multi-dimensional approach to disinformation: Final report of the High Level Expert Group on Fake News and Online Disinformation . https://doi.org/10.2759/0156 - · DOI
  12. [21][21] Miriam Fernández, Alejandro Bellogín, and Iván Cantador. 2021. Analysing the Effect of Recommendation Algorithms on the Amplification of Misinformation. arXiv:2103.14748 [cs] (March 2021). arXiv:2103.14748 [cs] - · arXiv
  13. [23][23] Aniko Hannak, Piotr Sapiezynski, Arash Molavi Kakhki, Balachander Krishnamurthy, David Lazer, Alan Mislove, and Christo Wilson. 2013. Measuring Personalization of Web Search. In Proc. of the 22nd International Conference on World Wide Web (WWW '13) . ACM, New York, NY, USA, 527-538. https://doi.org/10.1145/2488388.2488435 - · DOI · Paper page
  14. [24][24] Muhammad Haroon, Anshuman Chhabra, Xin Liu, Prasant Mohapatra, Zubair Shafiq, and Magdalena Wojcieszak. 2022. YouTube, The Great Radicalizer? Auditing and Mitigating Ideological Biases in YouTube Recommendations. arXiv:2203.10666 [cs] (March 2022). arXiv:2203.10666 [cs] - · arXiv
  15. [25][25] Jeff Horwitz. 2021. The Facebook Files: A Wall Street Journal Investigation. Wall Street Journal (Online) (2021). -
  16. [26][26] Rui Hou, Veronica Perez-Rosas, Stacy Loeb, and Rada Mihalcea. 2019. Towards Automatic Detection of Misinformation in Online Medical Videos. In 2019 International Conference on Multimodal Interaction (Suzhou, China) (ICMI '19) . Association for Computing Machinery, New York, NY, USA, 235-243. https://doi.org/10.1145/3340555.3353763 - · DOI
  17. [27][27] Eslam Hussein, Prerna Juneja, and Tanushree Mitra. 2020. Measuring Misinformation in Video Search Platforms: An Audit Study on YouTube. Proc. ACM Hum.-Comput. Interact. 4, CSCW1, Article 048 (May 2020), 27 pages. https: //doi.org/10.1145/3392854 - · DOI
  18. [29][29] Prerna Juneja and Tanushree Mitra. 2021. Auditing E-Commerce Platforms for Algorithmically Curated Vaccine Misinformation. In Proc. of the 2021 CHI Conference on Human Factors in Computing Systems (CHI '21) . https: //doi.org/10.1145/3411764.3445250 arXiv:2101.08419 - · DOI
  19. [31][31] Huyen Le, Andrew High, Raven Maragh, Timothy Havens, Brian Ekdale, and Zubair Shafiq. 2019. Measuring political personalization of Google news search. In Proc. of the World Wide Web Conference (WWW '19) . 2957-2963. https: //doi.org/10.1145/3308558.3312504 - · DOI
  20. [32][32] Ping Liu, Karthik Shivaram, Aron Culotta, Matthew A. Shapiro, and Mustafa Bilgic. 2021. The Interaction between Political Typology and Filter Bubbles in News Recommendation Algorithms. In Proceedings of the Web Conference 2021 (WWW'21) . Association for Computing Machinery, New York, NY, USA, 3791-3801. https://doi.org/10.1145/3442381. 3450113 - · DOI
  21. [34][34] Danaë Metaxa, Joon Sung Park, James A. Landay, and Jeff Hancock. 2019. Search Media and Elections: A Longitudinal Investigation of Political Search Results. Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 129 (Nov. 2019), 17 pages. https://doi.org/10.1145/3359231 - · DOI
  22. [35][35] Lien Michiels, Jens Leysen, Annelien Smets, and Bart Goethals. 2022. What Are Filter Bubbles Really? A Review of the Conceptual and Empirical Work. In Adjunct Proceedings of the 30th ACM Conference on User Modeling, Adaptation and Personalization (UMAP '22 Adjunct) . Association for Computing Machinery, New York, NY, USA, 274-279. https: //doi.org/10.1145/3511047.3538028 - · DOI
  23. [36][36] Diana C. Mutz and Lori Young. 2011. Communication and public opinion: Plus ça change? Public opinion quarterly 75, 5 (2011), 1018-1044. -
  24. [37][37] Richard Owen, Phil Macnaghten, and Jack Stilgoe. 2012. Responsible research and innovation: From science in society to science for society, with society. Science and Public Policy 39, 6 (2012), 751-760. https://doi.org/10.1093/scipol/scs093. - · DOI
  25. [38][38] Kostantinos Papadamou, Savvas Zannettou, Jeremy Blackburn, Emiliano De Cristofaro, Gianluca Stringhini, and Michael Sirivianos. 2022. "It Is Just a Flu": Assessing the Effect of Watch History on YouTube's Pseudoscientific Video Recommendations. Proceedings of the International AAAI Conference on Web and Social Media 16, 1 (May 2022), 723-734. https://ojs.aaai.org/index.php/ICWSM/article/view/19329 -
  26. [39][39] Eli Pariser. 2011. The filter bubble: What the Internet is hiding from you . Penguin UK. -
  27. [40][40] Branislav Pecher, Ivan Srba, Robert Moro, Matus Tomlein, and Maria Bielikova. 2021. FireAnt: Claim-Based Medical Misinformation Detection and Monitoring. In Proc. of the Joint European Conference on Machine Learning and Knowledge Discovery in Databases (ECML PKDD '20) . 555-559. https://doi.org/10.1007/978-3-030-67670-4\_38 - · DOI
  28. [42][42] Ronald E. Robertson, David Lazer, and Christo Wilson. 2018. Auditing the personalization and composition of politically-related search engine results pages. Proc. of the World Wide Web Conference (WWW '18) , 955-965. https: //doi.org/10.1145/3178876.3186143 - · DOI
  29. [43][43] Christian Sandvig, Kevin Hamilton, Karrie Karahalios, and Cedric Langbort. 2014. Auditing algorithms: Research methods for detecting discrimination on internet platforms. Data and discrimination: converting critical concerns into productive inquiry 22 (2014), 4349-4357. -
  30. [44][44] Hong Shen, Alicia Devos, Motahhare Eslami, and Kenneth Holstein. 2021. Everyday Algorithm Auditing: Understanding the Power of Everyday Users in Surfacing Harmful Algorithmic Behaviors. Proc. ACM Hum.-Comput. Interact 5, CSCW2 (Oct. 2021), 29. https://doi.org/10.1145/3479577 - · DOI · Paper page
  31. [45][45] Márcio Silva, Lucas Santos de Oliveira, Athanasios Andreou, Pedro Olmo Vaz de Melo, Oana Goga, and Fabricio Benevenuto. 2020. Facebook Ads Monitor: An Independent Auditing System for Political Ads on Facebook. In Proc. of The Web Conference (WWW '20) . ACM, New York, NY, USA, 224-234. https://doi.org/10.1145/3366423.3380109 - · DOI
  32. [46][46] Jakub Simko, Patrik Racsko, Matus Tomlein, Martina Hanakova, Robert Moro, and Maria Bielikova. 2021. A study of fake news reading and annotating in social media context. New Review of Hypermedia and Multimedia (2021), 1-31. -
  33. [47][47] Jakub Simko, Matus Tomlein, Branislav Pecher, Robert Moro, Ivan Srba, Elena Stefancova, Andrea Hrckova, Michal Kompan, Juraj Podrouzek, and Maria Bielikova. 2021. Towards Continuous Automatic Audits of Social Media Adaptive Behavior and Its Role in Misinformation Spreading. In Adjunct Proc. of the 29th ACM Conference on User Modeling, Adaptation and Personalization (UMAP '21) . ACM, New York, NY, USA, 411-414. https://doi.org/10.1145/3450614. 3463353 - · DOI
  34. [48][48] Larissa Spinelli and Mark Crovella. 2020. How YouTube Leads Privacy-Seeking Users Away from Reliable Information. In Adjunct Publication of the 28th ACM Conference on User Modeling, Adaptation and Personalization . ACM, New York, NY, USA, 244-251. https://doi.org/10.1145/3386392.3399566 - · DOI
  35. [49][49] Ivan Srba, Robert Moro, Daniela Chuda, Maria Bielikova, Jakub Simko, Jakub Sevcech, Daniela Chuda, Pavol Navrat, and Maria Bielikova. 2019. Monant: Universal and Extensible Platform for Monitoring, Detection and Mitigation of Antisocial Behavior. In Proc. of Workshop on Reducing Online Misinformation Exposure (ROME '19) . 1-7. -
  36. [51][51] Matus Tomlein, Branislav Pecher, Jakub Simko, Ivan Srba, Robert Moro, Elena Stefancova, Michal Kompan, Andrea Hrckova, Juraj Podrouzek, and Maria Bielikova. 2021. An Audit of Misinformation Filter Bubbles on YouTube: Bubble Bursting and Recent Behavior Changes. In Fifteenth ACM Conference on Recommender Systems . ACM, New York, NY, USA, 1-11. https://doi.org/10.1145/3460231.3474241 - · DOI · Paper page
  37. [53][53] Siva Vaidhyanathan. 2018. Antisocial media: How Facebook disconnects us and undermines democracy . Oxford University Press. -
  38. [57][57] Zhe Zhao, Lichan Hong, Li Wei, Jilin Chen, Aniruddh Nath, Shawn Andrews, Aditee Kumthekar, Maheswaran Sathiamoorthy, Xinyang Yi, and Ed Chi. 2019. Recommending what video to watch next: A multitask ranking system. In Proc. of the 13th ACM Conference on Recommender Systems (RecSys '19) . ACM, 43-51. https://doi.org/10.1145/ 3298689.3346997 -
  39. [58][58] Xinyi Zhou and Reza Zafarani. 2020. A Survey of Fake News: Fundamental Theories, Detection Methods, and Opportunities. Comput. Surveys 53, 5 (Dec. 2020). https://doi.org/10.1145/3395046 arXiv:1812.00315 - · DOI · Paper page
  40. [59][59] Fabiana Zollo, Petra Kralj Novak, Michela Del Vicario, Alessandro Bessi, Igor Mozetic, Antonio Scala, Guido Caldarelli, and Walter Quattrociocchi. 2015. Emotional Dynamics in the Age of Misinformation. CoRR (2015). http://dblp.unitrier.de/db/journals/corr/corr1505.html#ZolloNVBMSCQ15 -

Cited by

Sorted by citations

Authors
Srba, I., Moro, R., Tomlein, M., Pecher, B., Simko, J., Stefancova, E., Kompan, M., Hrckova, A., Podrouzek, J., Gavornik, A., & Bielikova, M.
Affiliations
  • Kempelen Institute of Intelligent Technologies (SK)
  • Kempelen Institute of Intelligent Technologies, Bratislava, Slovakia
  • Brno University of Technology (CZ)
ORCID
Venue
ACM Transactions on Recommender Systems
Publisher
Association for Computing Machinery (ACM)
Published
2023-01-27
Volume
1
Issue
1
Pages
1-33
ISSN
2770-6699 (ISSN-L)
DOI
10.1145/3568392
OpenAlex
W4307005348
Cited by
61
References
6
Citations by year
2026: 19 · 2025: 16 · 2024: 19 · 2023: 7
FWCI
13.30
Type
journal article
Language
en
Open access
Licence
www.acm.org/publications/policies/copyright_policy#Background (vor)
Locations
  • ACM Transactions on Recommender Systems (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Zenodo (CERN European Organization for Nuclear Research) (submittedVersion, cc-by)
  • arXiv (Cornell University)
Funders
  • Ministry of Education, Science, Research and Sport of the Slovak Republic (0827/2021)
  • Central European Digital Media Observatory
  • European Union (2020-EU-IA-0267)
  • EU Horizon 2020 (952215)
  • Horizon Europe, GA (101070093)
Topics
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Physical Sciences › Computer Science › Information Systems › Recommender Systems and Techniques
  • Physical Sciences › Computer Science › Information Systems › Spam and Phishing Detection
Keywords
YouTube, misinformation debunking, algorithmic bias, video classification, misinformation detection

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review