Algorithmic literacy is the term used to describe the capacity and opportunity to be aware of both the presence and impact of algorithmically driven systems (Oeldorf-Hirsch, 2023). Personalization can be defined as mechanisms to customize information and content to users (Perugini). Within personalization, building community is a key concept of TikTok’s aims to create permeable communities - communities one can weave in and out of - that allows users to discover not only their identities but others (Milton). This has its disadvantages as unregulated discoverability can showcase harmful content to users. However, personalization showcases an advantage to those who understand how to use the application, as they grasp the effects of curating their own experience, rather than just letting the algorithm adapt. Aside from algorithmic literacy, the curation element of TikTok is a reflection of navigation as the user has agency in what is presented to them - to an extent. With recommendation algorithms being based off of a user’s interaction such as search, should the For-You-Page reflect those searches is a question raised by Milton et al about the ‘feed’ of TikTok. There are concerns about whether or not searches are a reliable source of information about a user.

For marginalized users, content review and moderation is a large element of what their algorithmic experiences are. Shadow banning is a practice of removing videos without a user’s knowledge, which disproportionately impacts marginalized users (DelMonaco). Rong et al discusses how this tactic impacts blind users often suppressing their content to a more mainstream audience, and how they find themselves by trying to immerse themselves as much into staying in their desired communities (Smith, 2023).

To combat content suppression and mitigation there are a number of techniques that marginalized users partake in. A common method is avoiding hashtags as that spreads the material far out of the creators’ typical interactions and communities (DelMonaco). Another method is sticking to trending topics as there is a large saturation of material, so the content does not always spread as much.

BIPOC users are becoming more of a voice in bringing awareness to the harms of TikTok, both technically and socially. Williams conducted a study where Black creators shared content mitigation approaches to their interaction within the application. There is limited research on the nature of content BIPOC users encounter and how they interact with potentially problematic content. Some studies have examined specific content types such as when Hung et al (2023) conducted content analysis on videos tagged ‘#racialtrauma’. The team found a frequent theme of ‘questioning and challenging oppression, denial and privilege’, where creators drew comparisons of their own experience against other creators (Hung, 2021). This theme reflects on the experience of BIPOC individuals on TikTok, where they question and discuss their own experiences surrounding racism. Existing research has concentrated on content creators, particularly how BIPOC creators deal with content suppression and algorithmic bias. Although the data from this study was not discussed in the context of mental health or feelings surrounding content by BIPOC creators, it is important to note how BIPOC creators focus on education and advocacy through their content, to fight algorithmic bias.

Throughout all the literature discovered, there has yet to be a focus of BIPOC user experiences. There’s limited research on the nature of content BIPOC users encounter, especially on how they perceive and interact with potentially problematic content. While some studies have examined specific content types (e.g., Hung et al.’s study on ‘#racialtrauma’ videos), there’s a gap in understanding the broader content landscape for BIPOC users. A majority of the existing research has concentrated on content creators’ experiences, particularly how BIPOC creators deal with content suppression. There’s less information about the experiences of BIPOC individuals who primarily consume content on TikTok. Improving the TikTok experience is vital to creating an inclusive platform, as the neglect to amplify voices that should be seen reflects a failure to listen to those who deserve to be heard (Antoine, 2022).

The participants for the pilot study were selected based on specific inclusion criteria to ensure a diverse and relevant sample. We conducted recruitment by publishing a flyer, in the i3 Instagram and Slack, with the criteria titled ‘Research Opportunity: We Want to Hear About your TikTok Experience’. In the flyer we briefly explained the purpose of our study, inviting readers to join us in centring the voices of BIPOC users of TikTok. Eligibility criteria included:

BIPOC individual

The flier included a link to a Google form for potential participants to provide their contact information and confirm their eligibility. Emails were scheduled by giving each participant a few options for time slots. However, as we expand the number of participants, we will be using Calendly to make it easier for them to schedule time slots that work for our schedule as researchers. This flyer was available on i3’s Instagram and Slack channels.

Interviews were scheduled conveniently for participants and interviewers and conducted on Zoom. Upon conducting 10 semi-structured interviews, we transcribed the interviewers’ answers and put them in a Google Sheet File, split by question. Each interview lasted between 20 and 40 minutes. All the interviews were audio recorded with participants’ consent to ensure accurate transcription and analysis.

We transcribed the interviewers’ answers into an organized Google Sheet file, split by question. We fixed the transcription errors and only kept the complete answers from the participants, noting our follow-up questions. Then, each of us conducted qualitative coding, starting by writing out the initial themes for answers from all 10 participants. We included quotes that stood out to us in their answers, including a list of recurring themes. Then, we summarized those into focused codes - themes we’ve seen across multiple answers to the same question. Then, those focused codes were combined between answers to summarize findings for each question from the interview pool. For every response, we noticed similarities and themes across the questions to combine the topics and restructure the interviews. As is shown in Table 1, the age range of our participants varied between 19 and 27 instead of including adults in their 30s. This affected the scope of our study and only made the results applicable to the age group interviewed.

We examined the types of content BIPOC users encounter on their TikTok FYP (Table 2). The most frequently encountered content type was BIPOC Representation, followed by Humor and Entertainment, and Cultural Relevance. While this diversity suggests a broad range of content, the relative scarcity of educational or critical discourse content may indicate an algorithmic preference for entertainment over information.

Participants identified problematic content, including misinformation, discriminatory material, and unintentional stereotypes. Their responses included disengagement, algorithm manipulation, and information-seeking. However, the effectiveness of these responses remains uncertain.

Participants demonstrated an understanding of TikTok’s algorithm, emphasizing engagement-based recommendations, data collection, and cross-platform influence. While most accepted this as inherent to social media, concerns over privacy and content manipulation were evident.

BIPOC users actively curate their content by engaging with preferred material, avoiding unwanted content, and periodically resetting the algorithm. However, these strategies suggest a reactive rather than an empowered approach.

The high prevalence of entertaining (8/10 participants) and culturally relevant content (7/10 participants) on BIPOC users’ FYPs demonstrates TikTok’s effectiveness in curating personalized content, aligning with research on social media algorithms’ ability to create tailored content ecosystems (Eg et al., 2023). The significant presence of BIPOC-centered content (9/10 participants) suggests success in promoting diversity and addressing historical underrepresentation in mainstream media (Titanji et al., 2022). However, the emphasis on entertainment over education (4/10 participants) raises questions about meaningful knowledge dissemination.

All participants reported encountering problematic content, with a preference for disengagement (8/10) and overactive reporting (3/10). This behavior aligns with research on digital resilience strategies among marginalized communities (Du et al., 2020) and highlights ongoing content moderation challenges. However, this may not be effective in reducing exposure over time. Future studies could examine whether TikTok’s algorithm adjusts based on disengagement, or if problematic content persists despite user efforts.

BIPOC users demonstrate high algorithm awareness and actively attempt to influence their content recommendations, challenging notions of algorithmic determinism and emphasizing user agency in shaping digital experiences. The findings highlight an ongoing tension between personalization and control. While users employ strategies to influence content recommendations, they also express scepticism about TikTok’s neutrality. Theories of algorithmic bias (Noble, 2018) could provide a useful lens to further analyze these concerns.

This study included 10 participants, offering in-depth but limited perspectives. Future research should aim for a more diverse sample of BIPOC users, considering intersectional factors such as age, gender, socioeconomic background, and specific ethnic identities. This approach would provide a richer understanding of how different BIPOC subgroups interact with and perceive TikTok’s algorithm, acknowledging the complexity of intersecting identities and experiences