Selected Papers of #AoIR2023: The 24th Annual Conference of the Association of Internet Researchers Philadelphia, PA, USA / 18-21 Oct 2023 Suggested Citation (APA): Reynolds, C. and Hallinan, B. (2023, October). “YouTube Doesn’t Care About Creators”: How YouTubers Use the Platform to Promote Accountability. Paper presented at AoIR2023: The 24th Annual Conference of the Association of Internet Researchers. Philadelphia, PA, USA: AoIR. Retrieved from http://spir.aoir.org. “YOUTUBE DOESN’T CARE ABOUT CREATORS”: HOW YOUTUBERS USE THE PLATFORM TO PROMOTE ACCOUNTABILITY CJ Reynolds The Hebrew University of Jerusalem Blake Hallinan The Hebrew University of Jerusalem For platforms, moderation is, as Tarleton Gillespie argues, “essential, constitutional, definitional” (2018, p. 21). From removing copyrighted materials to targeting advertisements, from verifying accounts to creating revenue-sharing agreements, platforms make all kinds of decisions that set the conditions for who or what gets seen. But while platforms do extensive work regulating visibility on their sites, such moderation “must be largely disavowed, hidden” (ibid.). Unsatisfied with the black- boxing of algorithmic governance (Katzenbach and Ulbricht, 2019) and platform governance on YouTube more generally (Tarvin and Stanfill, 2022), creators have begun to seek accountability through other means, deploying their skills, audiences, and situated knowledge to investigate the platform’s operations. This paper examines a phenomenon we term user-generated accountability, or the use of publicity via content creation to reveal failures, oversights, or harmful policies on a platform. User-generated accountability is meant simultaneously to derive support from creators and audiences and to pressure the platform to acknowledge and remedy issues. Because creators are structurally incentivized to both care about and contest platform governance (Caplan and Gillespie, 2020; DeVos et al., 2022; Zeng and Kaye, 2022), creator perspectives and practices offer insight into the conditions of platformized cultural production and reveal novel political strategies. Furthermore, prior research has established that YouTubers actively engage with platform governance issues through content production, although studies tend to focus on specific issues such as copyright enforcement (Kaye and Gray, 2021) and demonetization (Caplan and Gillespie, 2020), or specific communities such as beauty vloggers (Bishop, 2019) and BreadTube (Cotter, 2022). Building on and broadening this work, we investigate YouTubers’ calls for accountability on the platform, asking the following questions: RQ1: What forms do calls for accountability on YouTube take? RQ2: What actors do YouTubers target when calling for accountability? RQ3: What topics or concerns do these calls address? Method To answer these questions, we analyzed videos featuring issues of platform accountability. In order to ensure our dataset featured diverse topics, concerns, and creators, we employed multiple sampling strategies. We started by creating “video networks” through YouTube Data Tools (Rieder, 2015), using a prominent trilogy of videos about a user-driven algorithmic audit of demonetization as the seeds (DeVos et al., 2022). We collected 94 videos from these networks and then ran 21 keyword searches using the “video list” function of YouTube Data Tools to collect an additional 1050 videos. We also manually supplemented the list, adding 185 videos to create a dataset of 1329 videos. After cleaning the data for duplicates and non-applicable videos, we had a corpus of n = 429 videos. To analyze the videos, we coded the format of the video, its stance toward YouTube, the targets of accountability, and the issues it raises. Following modified principles of the grounded theory approach (Kelle, 2007), which uses existing theoretical concerns to inform an inductive examination of the data, we developed the codebook starting from issues raised in prior research (Eslami et al., 2019; Haimson et al. 2021; Kingsley et al., 2022). After randomizing our corpus, we coded and engaged in constant comparison until both authors agreed that theoretical saturation had been reached at n = 250 videos. There are 172 unique channels represented in this sample. Findings Vlogs were by far the most popular format for accountability-focused videos, representing 165 cases or 66% of our sample. Scripted videos, such as video essays, were a distant second with 50 instances (20%), followed by conversation-style videos including podcasts and talk shows with 18 examples (7.2%). 14 videos in our sample were experiments meant to test the YouTube algorithm (5.6%), and only three videos fell outside these categories (1.2%). The overall stance towards YouTube in our data set was notably varied. We expected callout videos aiming for accountability to be overwhelmingly negative towards YouTube. Indeed, Negative was the largest category (83 cases or 33.2%), but Mixed, indicating both critique of and defenses of YouTube, appeared 73 times (29.2%). 58 videos took a Neutral/Undetermined stance towards YouTube (23.2%), while 36 videos primarily defended the platform (14.4%). Our accountability target codes (excepting the Not Applicable code) were non-exclusive. “YouTube” was cited the most often in videos that invoked at least one target of accountability, appearing in 141 videos (70.85%) from this subset. With all the talk of “the algorithm” among creators and scholars, Automation was the second-most invoked actor (55 cases, 27.64%). Surprisingly, “Self” was next with 46 cases (23.12%). Indeed, placing blame directly on human actors was frequent in our data set, with Other Creators (31 cases or 15.58%), YouTube Employees (27 cases or 13.57%), and YouTube CEO Susan Wojcicki (21 cases or 10.55%) easily surpassing targeting of automated systems when combined (79 cases, or 39.7% of accountability invocations). Only two other groups appeared more than 10 times in our sample, with Audiences invoked 21 times (10.55%) and Predators—those who exploit children or animals— targeted with considerable vitriol in 11 videos (5.53%), in line with previous research (Tarvin and Stanfill, 2022). We coded for 16 non-exclusive content elements within videos that identified specific complaints or problems with YouTube. Complaints about YouTube Policies appeared in over half our sample, 126 videos. Also prominent was anger about lack of communication (74 cases) and perceived bias against certain demographics of creators (72 cases) or types of content (67 cases). Censorship was claimed in 55 videos, and issues with YouTube features (45 cases), the corporation’s cultural disconnect with creators (32 cases), and the frustrating process for filing appeals (27 cases) all appeared in more than 10% of our sample. Conclusion User-generated accountability videos create a significant sense of solidarity between creators on YouTube, with many creators lending their voices to signal boost issues raised by others. In one video, 25 other YouTubers appeared on camera to back the grievances of creator MamaMax, whose video criticizing child predators had been removed by the platform. This is community building by both choice and necessity, as most creators expressed they felt they could not leave YouTube; no other platform offered them the reach that YouTube did, nor could they easily move their audiences elsewhere. Feeling trapped on YouTube led to significantly increased frustrations regarding the platform’s perceived lack of communication with its creators, to the extent that many reiterated the need to take to Twitter to force YouTube to respond to their appeals or policy disputes. If still unheeded, they did what they do best—make videos about the issue. Creators also highlighted how the lack of transparency about YouTube’s systems—algorithmic and human—contributed significantly to stress and burnout, even leading some to consider quitting content creation altogether. YouTube is a platform uniquely suited to user-generated accountability, as the entire structure of the site revolves around generating mass publicity and visibility. At the same time, however, all of YouTube’s problems are problems at scale due to their 2.6 billion monthly users. An issue that affects just 1% of the platform may not draw major corporate attention even as it plagues 26 million people. Thus, publicizing problems with the platform in a way that draws attention from audiences, news media, and other creators represents one of the most important ways YouTube’s creators can participate in platform governance. References Bishop S (2019) Managing visibility on YouTube through algorithmic gossip. New Media & Society 21(11–12): 2589–2606. DOI: 10.1177/1461444819854731. https://doi.org/10.1177/1461444819854731 Caplan R and Gillespie T (2020) Tiered Governance and Demonetization: The Shifting Terms of Labor and Compensation in the Platform Economy. Social Media + Society 6(2): 205630512093663. DOI: 10.1177/2056305120936636. Cotter K (2022) Practical knowledge of algorithms: The case of BreadTube. New Media & Society. DOI: 10.1177/14614448221081802. DeVos A, Dhabalia A, Shen H, et al. (2022) Toward User-Driven Algorithm Auditing: Investigating users’ strategies for uncovering harmful algorithmic behavior. In: CHI Conference on Human Factors in Computing Systems, New Orleans LA USA, 29 April 2022, pp. 1–19. ACM. DOI: 10.1145/3491102.3517441. Eslami M, Vaccaro K, Lee MK, et al. (2019) User Attitudes towards Algorithmic Opacity and Transparency in Online Reviewing Platforms. In: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, Glasgow Scotland Uk, 2 May 2019, pp. 1–14. ACM. DOI: 10.1145/3290605.3300724. Gillespie T (2018) Custodians of the Internet: Platforms, Content Moderation, and the Hidden Decisions That Shape Social Media. New Haven: Yale University Press. Haimson OL, Delmonaco D, Nie P, et al. (2021) Disproportionate removals and differing content moderation experiences for conservative, transgender, and Black social media users: Marginalization and moderation gray areas. Proceedings of the ACM on Human- Computer Interaction 5(CSCW2): 1–35. DOI: 10.1145/3479610. Katzenbach C and Ulbricht L (2019) Algorithmic governance. Internet Policy Review 8(4). DOI: 10.14763/2019.4.1424. Kaye DBV and Gray JE (2021) Copyright gossip: Exploring copyright opinions, theories, and strategies on YouTube. Social Media + Society 7(3): 205630512110369. DOI: 10.1177/20563051211036940. Kelle U (2007) The development of categories: Different approaches in grounded theory. In: Charmaz K (ed.) The SAGE handbook of grounded theory. London: Sage, pp. 191–213. Kingsley S, Sinha P, Wang C, et al. (2022) ‘Give everybody [..] a little bit more equity’: Content creator perspectives and responses to the algorithmic demonetization of content associated with disadvantaged groups. Proceedings of the ACM on Human- Computer Interaction 6(CSCW2): 1–37. DOI: 10.1145/3555149. Rieder B, (2015) YouTube Data Tools (Version 1.30) [Software]. Available from https://tools.digitalmethods.net/netvizz/youtube/ Tarvin E and Stanfill M (2022) “YouTube’s predator problem”: Platform moderation as governance-washing, and user resistance. Convergence: The International Journal of Research into New Media Technologies 28(3): 822–837. DOI: 10.1177/13548565211066490. https://doi.org/10.1177/2056305120936636 https://doi.org/10.1177/14614448221081802 https://doi.org/10.1145/3491102.3517441 https://doi.org/10.1145/3290605.3300724 https://doi.org/10.1145/3479610 https://doi.org/10.14763/2019.4.1424 https://doi.org/10.1177/20563051211036940 https://doi.org/10.1145/3555149 https://tools.digitalmethods.net/netvizz/youtube/ Zeng J and Kaye DBV (2022) From content moderation to visibility moderation: A case study of platform governance on TikTok. Policy & Internet 14(1): 79–95. DOI: 10.1002/poi3.287. https://doi.org/10.1002/poi3.287