Algorithmic imaginings and critical digital literacy on #BookTok

Abstract

Despite the growing impact of algorithms on digital culture, and the importance of algorithmic awareness, little literacy research has investigated how algorithmic awareness and speculation shapes cultural production on digital platforms. Developing Bucher’s concept of the “algorithmic imagination” for digital literacy research, we conduct a study of #BookTok, the home of book-related content on TikTok, the most algorithm-driven social media platform to date. Through a multimodal content analysis of 57 videos containing #algorithm and #BookTok, we propose and explore a typology of five categories of “algorithmic imaginings”: critique, defense, explanation, how to work, and exploration of the algorithm. These imaginaries move beyond rational attempts to deconstruct the algorithm and critique its role in platform capitalism toward playful explorations of the human–algorithmic relationship. This constitutes for us another dimension of critical literacy, as producers anthropomorphize technology in a manner that addresses the symbiotic meaning-making of human and machine head-on.

The datafication of online social life has forced literacy researchers to think differently about the nature of digital literacy as a rational, human practice. Earlier in the 21st century, literacy researchers considered digital literacy in terms of learning multimodal discourses and skills necessary to engage in online, participatory culture, which opened opportunities for civic engagement, fandom, and cultural production. Now, proprietary algorithms constantly refine and co-produce culture alongside users, shifting cultural production from a fully human activity to a human–artificial intelligence (AI) co-production. This shift in understanding online contexts of literacy from participatory to algorithmic cultures has been forced in part by platforms’ proprietary algorithms that corral users into pens of limited and insular discourses (Benjamin, 2019). This algorithmic co-determination of participation and cultural formation has placed new demands on digital literacy, including the necessity of critical, “algorithmic awareness” (Hargittai et al., 2020). Yet, little literacy research has investigated how humans’ algorithmic awareness affects their activity online. How does one become critically literate in the practices and discourses of cultures that are co-created with what are largely black-boxed algorithms (Dogruel et al., 2022), the workings of which users can only partially understand through extended use (Cotter and Reisdorff, 2020), and trial and error?

In this article, we argue that research on critical digital literacy must also account for how algorithmic awareness transforms online production into cultural (co-) production between humans and machines. We use the term cultural production (or co-production) to signal the ways those creating, posting, and engaging with videos on TikTok are also participating in its algorithmically driven social processes—consequently generating, curating, and circulating #BookTok and TikTok culture. In so doing, we develop the concept of “algorithmic imaginaries” for digital literacy research.

An important context for this work is “critical algorithm studies,” which explore the social impacts of algorithms, including algorithmic biases, surveillance and discrimination (Benjamin, 2019; Noble, 2018), governance systems (Gillespie, 2022), and the “political ramifications” of their “knowledge logic” (Gillespie, 2013). Most relevant to our work are empirical studies on user awareness of and attitudes toward algorithms and their implications for “public participation and democracy” (Gran et al., 2021) as well as information literacy. These examine algorithmic awareness across populations, and document correlations between greater understandings and education levels (Dogruel et al., 2022), socio-economic status (Cotter and Reisdorff, 2020), and ability to code, leading to concerns about deepening digital divides (Gran et al., 2021).

Through this work, scholars are developing a conceptual vocabulary to discuss how people relate to algorithms. Bucher (2017) describes “algorithmic imaginaries” as “the way in which people imagine, perceive and experience algorithms and what these imaginations make possible” (p. 31). Algorithmic imaginaries reveal how cultural producers understand, engage, and talk about a social media algorithm that they need to “work” to make their content more visible and to better curate their own experiences on a platform. They resemble “algorithmic gossip,” the “communally and socially informed theories and strategies pertaining to recommender algorithms, shared and implemented to engender financial consistency and visibility” (Bishop, 2019: 2589). They also relate to “folk theories,” the “intuitive, informal theories that individuals develop to explain the outcomes, effects, or consequences of technological systems” (DeVito et al., 2017). In this article, we build on the concept of the algorithmic imaginary in particular because it signals the speculative quality of this popular theorizing, since there is no definitive knowledge of how algorithms work, given their proprietary and changing nature, as well as their indeterminacy; as with AI more generally, engineers confess that they do not fully know or determine how the algorithms they build operate (Knight, 2017). In Gran et al.’s (2021) list of important directions for algorithm studies, they recommend “studying algorithm awareness and attitudes as part of Internet cultures” (p. 1792); our study of “algorithmic imaginaries” takes up this call. We will also explain how these imaginaries move beyond rational attempts to explain the algorithm and to critique its role in platform capitalism, which are more common approaches for critical literacies (see Nichols and Garcia, 2022).

This argument is developed through an analysis of how the algorithmic imagination shapes users’ cultural production on TikTok, ¹ a short-form video creation and sharing platform, focusing on one TikTok subculture: #BookTok, ² the home of book-related content, including book reviews and recommendations, trend critiques and commentaries, unboxing videos, “shelfies” (“bookshelf selfies”), and TBRs (To-Be-Read books). The #BookTok community is dynamic and growing, more than tripling in views during a few months of the pandemic, “from 3.4 billion views in February 2021 to 10.3 billion in June 2021” (Wiederhold, 2022: 157). This growth seems to reflect a renewed interest in reading more generally among young people. Publishers have taken note and now regularly collaborate with #BookTokers in marketing their work, and booksellers often feature “as seen on #BookTok” collections. As with BookTube, #BookTok’s YouTube counterpart, the most popular genre by far is Young Adult (YA) Literature. As a space tied to, and reportedly nurturing, Gen Z’s out-of-school reading practices, #BookTok is commanding attention among educators and literacy researchers, as well as librarians (Jerasa and Boffone, 2021; Merga, 2021; Roberts, 2021). We describe how #BookTok content producers, mostly youth (Vogels et al., 2022), imagine algorithmic operations in relation to the content they produce about their reading lives, asking, How does the algorithmic imagination shape cultural production on #BookTok?, and What are the implications of this for critical digital literacy?

TikTok as algorithmic culture

TikTok is synonymous with Gen Z (aged 10–25), said to comprise a major (60%) and growing segment of its viewership (Wiederhold, 2022). It is also the most algorithmically driven of the social media platforms (Bhandari and Bimo, 2022) to date.

The algorithm’s influence is most evident in each user’s For You Page (FYP), which offers a personalized content stream of videos. TikTok creators “strive to have individual posts accumulate ‘engagements’ in the form of views, comments, and shares” (Abidin, 2021: 79) to maximize their visibility on FYPs. This requires actively seeking to understand and work the algorithm by, for instance, encouraging viewer engagement and reproducing the most popular types of content. To get and keep attention, creators must be alert and nimble, “actively and very quickly adapting from the latest trends and viral practices on TikTok, to attempt varieties of styles—across hashtags, keywords, filters, audio memes, narrative memes” (Abidin, 2021: 79). Memetic audio-clips, in particular, are considered “the driving template and organizing principle of TikTok” (Abidin, 2021: 80). These adaptive strategies participate in what Merga (2021) calls the “memetic behaviour” central to TikTok: its “celebration of appropriation, repetition and imitation” (p. 2). Users also try to influence the algorithm to curate (and so improve) their viewing experience. Their algorithmic practices include liking, commenting upon, saving, and sharing posts in various TikTok subcultures, driven by specific interests and “aesthetics,” “the Gen Z-friendly term for highly stylized visual trends” (Weekman, 2020), such as #cottagecore, #darkacademia, and #egirl.

The algorithm is so central to TikTok that there is a tradition of video reflections (as well as Reddit threads and tech journalism) on its workings (Schellewald, 2021). In a move to greater transparency, TikTok has published basic information about the “recommendation system” that shapes For You feeds, indicating that it is based on “user interactions” with content, “video information,” and “device and account settings” such as country and language (https://newsroom.tiktok.com); however, this information does not seem to have lessened attempts to uncover the algorithm’s more specific mechanisms. TikTok’s algorithms are an example of “platform logics” that shape its “content creation processes” (Kaye et al., 2021) as well as the “communicative forms,” which are “the platform-specific languages or memes, trends, and aesthetic styles that are specific to TikTok and the meaning-making practices of its users” (Schellewald, 2021: 1439).

The human–algorithm debate: who is the cultural producer?

Throughout the scholarship, algorithms are regarded as somewhat opaque curatorial devices (Swart, 2021); this opacity has not stopped researchers and end-users alike from trying to make sense of these dynamic yet invisible agents swaying human digital practices and behaviors (Eslami et al., 2016; Rader and Gray, 2015). Their dynamicity is part of what makes them invisible—or rather, forgettable—for most users. It is mainly when something is amiss that users notice the presence of algorithms (Lomborg and Kapsch, 2019). For example, deriving from Carrington’s (2018) notion of “algorithmic identities,” social media users mainly take notice when targeted by advertising or content that they feel does not mesh with their self-crafted identity (or aesthetic). Whether or not users are aware of the presence and workings of the algorithm, the human–AI relationship is constant. Siles et al. (2019) describe this relationship as a continuous and interdependent cycle of “mutual domestication.” In particular, personalization tactics have contributed to this dynamic, with machine learning optimizing the overall platform experience by compiling and analyzing user data—consequently promoting content that best resonates with each individual user (Carrington, 2018; Siles et al., 2019). The relationship between creator and algorithm is so central to TikTok that Bhandari and Bimo (2022) describe self-representation on the platform as creating an “algorithmized self,” as users “engage with versions of themselves, as mediated through the algorithm” (p. 9). The concept of the human–algorithmic relationship as a symbiotic cycle guides this article, wherein humans and algorithms alike process inputs from the other, generating outputs which will influence both human and algorithmic behaviors accordingly. Neither human nor algorithm is ideologically neutral.

Critical digital literacy in algorithmic cultures

Contemporary conceptions of critical digital literacy acknowledge that literacy is multimodal and shaped by cultural values (Pahl and Rowsell, 2020). That is, as users consume and produce texts in digital cultures like #BookTok, for example, they intentionally orchestrate multiple modes, such as sound, image, and text, to create videos (and hashtags) that are recognizable to other #BookTokers as part of a culture. Literacy is also shaped by ideologies, which flow through this process of textual consumption and production in myriad ways: from how book publishers influence videos’ content to how producers can convey cultural stereotypes in representing books’ contents (Paladines and Aliagas, 2023). Part of becoming critically literate on #BookTok involves developing a critical consciousness attuned to how these ideologies are represented and circulated through texts (including novels popular on #Booktok, and the #BookTok videos themselves) and to how the production of new texts may counter such ideologies (Luke, 2018).

Platform logics have complicated this perspective on critical literacy (Nichols et al., 2022). Ideologies mobilize online not just in how they are represented semiotically in texts but also through how platforms like TikTok structure interactions. For example, algorithms that carry the social biases and capitalistic interests of their designers influence what texts are seen by whom. Most current conceptions of critical digital literacy therefore involve a consciousness of how platforms operate as organizing powers in digital social life (Garcia and de Roock, 2021). However, these conceptions of critical digital literacy remain focused on a rational, human-centered critique of representations and the ways ideologically driven platforms affect their circulation alongside users. This fails to capture fully the human–machine co-production of algorithmic cultures and its effect on human meaning-making.

Since the objects of interest in #BookTok are books, their authors and publishers, and book culture more generally, this forces the question of how algorithms inform meaning-making in our reading lives. Furthermore, as users experiment and speculate as to what videos may get pushed into other #BookTokers’ feeds, they are imagining what others and an algorithm may value in their reading and representation of a specific book. This process of production in algorithmic cultures suggests that critical digital literacy may need to move beyond critique of representations and platform logics alone (although these dimensions remain essential) to also include critical consciousness in algorithmic cultures.

#BookTok and the “digital literary sphere”

A precursor to #BookTok is the book blog, emerging in the late 1990s and early 2000s (Driscoll, 2019), developing into the book vlog (video log) dominating BookTube, Bookstagram, and #BookTok. Like book vlogs, such blogs were integrated into the fabric of the “reading industry” (Fuller and Rehburg Sedo, 2013) and performed “how to read, as well as what to read” (Driscoll, 2019: 301). Both #BookTok and book blogs exist as part of what Murray (2015) calls the “digital literary sphere,” at the intersection of literary culture and new media ecologies. This sphere is shaped by publishers, authors, readers, and booksellers, in communication—or “book chat” (Driscoll, 2019)—through “user-generated forums” (Murray, 2015: 324) and other channels, creating a “contemporary culture of reading” (Fuller & Rehburg Sedo, 2013).

Given that publisher websites and many user forums are corporate-owned (e.g. Goodreads is now owned by Amazon), with complex terms of use giving copyright control to the publishers, reader engagement in the digital literary sphere is enmeshed in the publishing industry, with participation producing free marketing content for corporations (e.g. reviews, fan fiction, fan art, author interviews) (Murray, 2015). Martens (2011) describes how young people participating on publisher sites can be branded and then marketed to through their identification with books, “targeted as both consumers and creators of the cultural products created for them” (p. 50).

Since a study of #BookTok is an exploration of youth engagement in contemporary book/reading culture as well as algorithmic culture, it raises questions about the place of algorithmic literacy in the digital literary sphere, making it of interest to literary studies and librarians, as well as digital and critical literacy researchers, scholars of new media, and educators.

Methods: researching #BookTok and the algorithmic imagination

Multimodal content analysis

To build our data corpus, we searched for TikTok videos that contained two specific hashtags, #algorithm and #BookTok; this approach is a form of “directed content analysis” which allows the researchers to identify “relevant work” since hashtags are such a “significant descriptive feature” on TikTok (Merga, 2021: 3). At the time of data collection, TikTok’s search function yielded a total of 57 videos using both hashtags. Beyond that, search results branched off thematically as TikTok’s algorithm attempted to diversify our feed (e.g. by recommending videos from “content coaches,” discussing and disseminating their algorithmic folk theories). The TikTok algorithm presents challenges for researchers attempting to document breadth of content because each person’s viewing is shaped by the recommendation engine shaping the FYP. Researchers have attempted to disrupt the algorithm, creating new accounts (Merga, 2021), or crafting two profiles and either actively engaging or resisting engagement as much as possible (Zulli and Zulli, 2020). We do not seek to make definitive claims about #BookTok content, which is always changing, but instead want to start building a data corpus exploring the algorithmic imagination’s impact on #BookTok cultural production.

We then undertook a multimodal content analysis of the data. Content analysis is a “systematic, rigorous approach to analyzing documents obtained or generated in the course of research” originally developed in communication studies (White and Marsh, 2006: 22) that seeks to draw inferences from texts to contexts in relation to specific research questions or hypotheses (Krippendorff, 2004). Qualitative content analysis is driven by open research questions, and the researcher must remain attuned to the unanticipated emergence of “concepts and patterns” and thus to revising initial questions (White and Marsh, 2006: 34), as well as to the possibility of multiple interpretations (p. 36). Approaching qualitative content analysis multimodally involves exploring “the complex relationships among modes present in analogue and digital multimodal ensembles, in addition to the affordances and limitations of individual modes” (Serafini and Reid, 2019: 7). This includes examining the videos as more than the sum of their parts—how they have been tagged, the interaction of audio and visual elements, including embedded text and TikTok features such as filters—to understand the multilayered meanings of these texts, including authorial intent. We also use quantitative analysis to document some of the forms and content features of the data corpus, itemizing key elements of each video under analysis, which allowed us to note how often certain ones occurred (Merga, 2021).

Each of the 57 videos was itemized with information about the content producer and a brief description of their content. First, quantitative data were collected per user account and specific #BookTok post, to give a sense of the reach and popularity of the account and content, respectively. Account popularity, reach, and dynamics were loosely judged based on: (1) the number of followers (higher numbers corresponding with greater influence), (2) number of following (a disproportionate follower-to-following ratio indicates the influential dynamic of the account, and a less personal relationship with followers), and (3) number of “likes” the account had accumulated (total sum of all posts, another indicator of popularity, and postreception by followers). Some creators shared demographic information, such as country, gender and sexual identity, and personal pronouns in their profiles; when this information has not been specified, we leave it out of our analysis. Moreover, we noted whether the account in question belonged to an author—those using #BookTok to actively self-promote or to sell their work. Further analysis of videos focused on the type of audio employed, for example, using original audio commentary versus participating in a trend by means of memetic audio clips. We also noted some of the uses of sound, such as using a song to heighten emotion. Videos using trending audios were usually humorous in nature, for example, playfully critiquing or expressing a quick opinion on a popular book or #BookTok or the TikTok algorithm. Some “talking” videos had original audio commentary from the producer on a book and/or author, #BookTok culture, the TikTok algorithm, or some combination of these. The type of audio used helped establish the tone of the video, and thus the purpose of the post.

We then took an inductive analytic approach to the data by coding the content of the videos regarding how the producers addressed or related to the algorithm, including a recursive process of open coding (involving initial loose and tentative coding of various approaches to the algorithm) and axial coding (in which connections were made between the codes, and codes were refined, combined or deleted) which lead to the development of categories; these emerging relationships and categories were refined and retested in relation to the corpus. Two of the researchers were involved in the initial open coding. Then, one researcher drew upon both sets of codes in the category development stage. All three researchers reviewed and validated the final categories.

Intergenerational researcher positionality

Our multimodal content analysis is also shaped by our previous studies and experiences of online book-related and other communities (Ehret, 2018), which are somewhat generationally shaped. Given the idiosyncratic nature of TikTok in particular, the project benefited from Anita, a member of Gen Z, TikTok’s predominant user base. (In contrast, Christian is a Millennial, whereas Bronwen is a member of Gen X.) Anita, who describes herself as “terminally online” (Hagh, 2020), spent countless hours on TikTok and was a cultural insider during the analysis, often responding to the research team’s questions about the platform’s dynamics, evolution, and trends. This intergenerational discursive dynamic during research term meetings created a constant refinement and member-checking of our interpretations of TikTok culture.

Even so, the ever-changing nature of the TikTok algorithm still posed problems. For example, during the data collection stage, videos could not be saved via “static” permalinks since TikTok employs a content distribution network that loads data dynamically; this meant that video links were prone to reshuffling, rendering saved links “broken” after some time. Also, our findings should be read as a snapshot in time, since the top recommended videos change frequently, as do other identifiers such as likes; the data collected reflects its state on the day it was sampled. ³

Rethinking critical digital literacy via #BookTokers’ algorithmic imaginings

Our sample of 57 videos includes videos by 34 different content creators, some of who created multiple videos. Of the 34 content creators in our sample, 22 creators noted which pronouns they use—17 indicated she/her (and one they/she or he), whereas only 4 indicated he/him pronouns, suggesting that predominantly female-identifying producers are exploring the algorithmic imagination on #BookTok. The only creators who noted their sexual or gender identity were not heterosexual (see Boffone and Jerasa, 2021, on how TikTok is used to support queer YA reading communities). Not all demographic information is available about every video or creator since sharing this is voluntary. Much of the content produced and released was by those who identify themselves as authors (n = 12). Authors were mostly concerned with marketing their novels and works (rather than, for example, discussing some popular books with fellow #BookTokers), and since the platform changes so quickly, they described needing to continually update their engagement strategies.

In response to our question of how does the algorithmic imagination shape cultural production on #BookTok?, we developed a typology of five categories of what we are calling “algorithmic imaginings.” Each video in the corpus fell within at least one category, and a few related to two categories, meaning some have been counted more than once. These categories are Critique of the Algorithm (n = 15), Defense of the Algorithm/Critique of Users (n = 6), Theorizing/Explanation of the Algorithm (n = 6) (akin to DeVito et al.’s (2017) “folk theories”), How- to Work the Algorithm (n = 16), and Exploration of the Algorithm (n = 15). In response to our second question, what are the implications of this for critical digital literacy?, our analysis illustrates the range of ways the algorithmic imagination shapes critical literacies on #BookTok, from the more traditional, such as resisting corporate influence in the Critique of the Algorithm theme, to the more novel Exploration of the Algorithm, which includes developing empathy toward a non-human being, the imagined algorithm, who represents these corporate influences. Across sections, we therefore address our research questions and raise new avenues for research on critical digital literacies in algorithmic cultures, highlighted in “Discussion and implications”.

Critique of the algorithm

It should not surprise that modes of critique were the most frequent form of algorithmic imaginary given that users tend to notice the algorithms when they interfere with their work (Lomborg and Kapsch, 2019). Critiques of the algorithm all offered critical commentary on the TikTok algorithm, and sometimes on the algorithm as specific to #BookTok. The tone varied between sincere/serious concern and humorous complaints. In an example of the former, @listenwithbritt noted that the “new TikTok update makes no sense” since the interface now refers to people by their name rather than more commonly known username, and then outlines multiple reasons this is a problem. In the comedic vein, @thesmutfairy is shown, looking bemused, in front of a screenshot of her posts (some viewed by over 2000 people, with others less than 300), with the titling “does the new tiktok algorithm hate booktok,” and to the audio meme “ok, batman, we’ll take it from here.” Some critiques came from authors concerned about the algorithm’s impact on their views and hence their business. @authoralexandrialee asks, “can someone explain the new tiktok algorithm?,” complaining about declining book sales because videos that used to get 2000–3000 minimum views now do not seem to attract beyond 200–300.

These critique videos often showcase the expertise of their creators (e.g. @listenwithbritt’s mention of “what I have learned over my ten years on social media”). Some include appeals for continued support despite the biases of the algorithm (e.g. @thesmutfairy, “Make sure you are interacting with your favorite creators. The struggle is real”). Since the critiques of the algorithm are also critiques of TikTok more generally, given the algorithm’s centrality to the platform’s character and operations, they are a familiar form of critical literacy, as users seek to explain and criticize the operations of corporate-controlled media platforms. They speak to a frustration with the algorithm’s control over what gets viewed and suggest that a way to “fight back” is to publicly speak out against the algorithm. The feelings of powerlessness are particularly troublesome for authors, according to our sample, since they use TikTok to promote their work, and for those committed to exploring and supporting marginalized YA texts and communities rather than already popular, trending, and often more mainstream texts. Systemic bias in the publishing and reading industry has been well-documented as part of “white literary taste production” operating at multiple levels, including reviews and prizes (Dane, 2023). #BookTok algorithms generally reproduce the industry’s focus on white, cisgendered, and able-bodied authors by promoting the same authors (e.g. Colleen Hoover, Sally Rooney). However, TikTok’s increasingly large and diverse user base means that many participants use #BookTok to actively amplify marginalized voices in the book world.

Such critiques of algorithmic culture echo those of other theorists documenting how AI encodes racist and sexist bias, as in Noble’s (2018) work on search engines, Nakamura (2019) on how AI can discriminate against those with disabilities, and articles in Benjamin’s (2019) collection on AI as surveillance technology. As Striphas (2023) notes, algorithmic bias is “a feature, not a bug” (p. 16).

Defense of the algorithm/critique of users

Defenses of the algorithm, significantly less frequent than critiques, were all responses to critiques from users such as those in the above section. They most often employed sarcasm, counterargument, and humor to counter their claims and concerns. For instance, @lin_reads takes issue with those who complain that they only ever see #BookToks about popular authors Sarah J. Maas and Colleen Hoover, arguing that the algorithm only reflects the videos that you watch and like. She appeals to viewers to follow more diverse #BookTok creators, noting in a snarky voice, “reading other books is not a flex” and “Please find your own niche and don’t make others feel bad about reading.”

@claudia.lynette’s post has a more sincere and serious tone and begins by listing some critiques of the algorithm (e.g. “booktok is dead” and “I only see the same 5 books”) and then shows the book covers of a collection of diverse texts on “my FYP recently,” noting that “the algorithm knows what it’s doing.”

Interestingly, defenses of the algorithm always go hand in hand with critiquing fellow #Booktokers. For users like @claudia.lynette, it is hard to be critical of the algorithm when the algorithm is your “friend.” Critical digital literacy is not as simple as critiquing platform logics if the platform logics work for you, as you work with them in the form of a friendly anthropomorphized machine.

Theorizing/explanation of the algorithm

Explanation imaginings draw upon folk theorizations (DeVito et al., 2017) of the algorithm and how it changes. For example, @baileyeliza shares a series of popular theories about how the algorithm works but also notes that these are “conspiracy theories” since she knows “next to nothing”; similarly, she describes her video as featuring “algorithm tips” but then says “I don’t claim to know if any of this is true.” Sometimes users correct each other's theorizations. For instance, @themusicaldealer comments, and then looks bewildered: “When you learn from your dad—an Internet marketing teacher—that TT’s algorithm is SO STRONG that it can recognize books in your background of your videos and will show your content to people who also like those books.” This form of sensationalism drives traffic but it also feeds into users’ perceptions of the algorithm as something mystical and frighteningly sophisticated. Yet, as two comments make clear, this theory is probably wrong: “I work in tech did my thesis on image recognition. the odds of your dad being right are next to 0” and “the machine learning and computer vision required for that process are way beyond what tiktok has the capacity for. it’s actually highly complicated.” Through such responses, users attempt to curb misinformation, which could push back against the platform’s capitalistic interest in driving traffic (without regard for the validity of the posts’ algorithmic theorizations). Critical literacy in this instance is not just about discerning truth but can also about involve countering the platform’s interests.

Exploratory imaginings often include more than a rational focus on working the algorithm, extending into how it might make users feel. For example, when @booklovingmisfit claims in a “new feature alert” that TikTok will now show who views your video, they mention that it could increase anxiety (i.e. “because we already worry about not enough views, now we will take it personally when someone [e.g. a friend] views but doesn’t like the video”). Critical literacy in this category is more than rational, including an awareness of how platform logics and algorithms can impact personal relationships.

How to work the algorithm

While similar to theorizing/explanation, how-to-work imaginings employ a more authoritative, knowing tone, building claims to expertise by educating fellow users on how to succeed at “working the algorithm” (i.e. getting the algorithm to promote one’s posts and/or show you posts optimized for one’s tastes). For instance, see @readwithloz, who devoted an entire playlist to instructional (how-to) videos of how-to-work the algorithm, signaling confidence and authority to her fellow #Booktokers. How-to-work videos were often used by authors for professional purposes and tended to have a sincere and serious tone, as in the six videos directed at fellow authors seeking to market their books on #BookTok. In this sense, the how-to-work imagining resembles Bishop’s (2019) description of the strategic aspects of “algorithmic gossip,” “shared and implemented to engender financial consistency and visibility on algorithmically structured social media platforms” (p. 2589).

@bonreviewsbooks describes a winning format: singing videos with storytelling and pairing these with trending audio. The video also suggests that using trending audio as a soundtrack, volume turned low, for “talking” TikToks can help the algorithm pick it up. It also posits that there is no correlation between views and likes, and, slipping into critique, notes that the algorithm does not pick up talk about books that are not already popular. Sharing their knowledge builds #BookTokers’ authority as #BookTok literate, and can thus be a tactic to gain more followers: @crystalandfelicity say, “follow us for more tips! the algorithm is NOT everything but this right here is REALLY important.” However, despite the confident tone in some of these videos, both explanations and how-to imaginings remain speculative, given the opacity of algorithmic workings. @jasondoroughauthor’s parody of the how-to genre speaks to this mystery: the creator says he has “figured the TikTok algorithm out: make sure fans watch all the way through” and that the “only way to do that is—” and then the video abruptly ends.

Both the explanation and how-to algorithmic imaginings are a collaborative form of critical literacy, as creators share what they have learned with other users and creators. They build community through recognition of a shared struggle against the seeming whims or biases of the algorithm, as well as strategies for combating these. Users are not just imagining how algorithms work: they are also working the algorithm in anticipation of, and in coordination with, other users working the algorithm. Culture itself is not just co-produced alongside an algorithm, but it is strengthened through a form of critical literacy that unites users in shared algorithmic imaginings.

Exploration of the algorithm

Exploration of the algorithm is typically signaled by the most common strategy, role-play (n = 8), in which the creator takes on various platform personas, such as the algorithm, the FYP, or the #BookTok community. While role-play is the most typical form of exploration, other notable versions include surprise and the recommendation journey. Despite being the most diverse video type, all videos in the sample used strategies of humor, including parody and sarcasm. “Exploration” is by far the most popular (as judged by likes) type of algorithmic imagining in our sample, with the top three most-liked videos all falling into this category. (At the time of data collection, the most popular video in our sample, by @paperbackprince, generated 90.5 K, followed by the next closest, @munnybooks with 30 K and @jamie.books 28.7 K.)

As examples of this embodiment of platform personas, one creator performed being the algorithm offering book suggestions to a picky user; another enacted a conflict between the algorithm and FYPs. In the first, @ezeekat, poking fun at the algorithm, initially embodies “BookTok” (as indicated in an embedded caption), reading Can’t Take That Away by Salvatore (which has a genderqueer teen protagonist); the creator then role-plays “The Algorithm,” holding up a piece of paper with “Enemies to Lovers,” a common YA theme, and then “With Spice,” a YA sexiness metric. In response, “BookTok” claps its hands and says “yes,” seemingly agreeing with the algorithm’s choices. In the second, @munnyreads also personifies the algorithm (again noted via embedded text) in conversation with the FYP. The algorithm suggests a series of books, to which the FYP says no and shakes its head, until the algorithm, resignedly, recommends Sarah J. Mass’s A Crown of Thorns and Roses, which gets a “wow” and a hand clap from the FYP. This book, sometimes referred to as ACOTAR, is both a major bestseller and a frequent target of scorn on #BookTok, perhaps because of what some have called “major consent and sexual abuse issues” (Tiff, 2015). The video’s borrowed audio can also be read as a commentary on the TikTok FYP algorithm, as well as users’ understandings and uses of intertextuality and multimodality. The sound originates from a #BookTok by Portuguese poster @pricilaagnes, in which the creator complains about the algorithm’s promotion of their seemingly least thoughtful comment. The original audio notes how “[doing a] little dance” or “asking for a cookie” (an expression used to indicate asking for attention) prompts an uninterested “no” from TikTok. Meanwhile, “me giving up and posting any sh*t” elicits a “woooooww” and much clapping from TikTok’s FYP/audience.

Another strategy for performing TikTok “insiderness” (literacy) is the use and exploration of #BookTok “tropes,” a common concept for readers and fans of YA literature, as well as fanfiction readers and writers. A trope can be a frequent theme, cliché, relationship dynamic, plot pattern, or character type (and thus has a broader meaning than the more typical literary use of trope as a figure of speech). For instance, in an enactment that begins with the following text, “POV: tiktok is trying to figure out your booktok algorithm,” @bookslumper roleplays both TikTok and The Reader, the former asking the latter a series of questions to which The Reader repeatedly says no: What kind of reader are you? Do you like sad endings? Miscommunication tropes? Shapeshifting romance? Love triangles? Pregnancy tropes? Stockholm spice? These questions about dominant #BookTok tropes are shared via embedded text, to the background of trending audio from Wreck-it Ralph. The video ends with The Reader asking: “are you guys ok? should I call the police?”

@bookslumper’s roleplay is a critical exploration of the imagined #BookTok algorithm, commenting not only on how she perceives YA Lit to be trope-driven but also on how she imagines the #BookTok algorithm reducing human reading interests to these tropes. The video’s ending suggests being overwhelmed by the imagined algorithm’s flood of tropes as well as feeling misunderstood and having her interests reduced to tropes. This sense of having human interests complexity diminished by a machine’s guesses is an impetus for critical literacy on #BookTok and for @bookslumper. @bookslumper both critiques a literary genre, YA, and how algorithmic culture feels reductive to her in its in capacity to fully represent and anticipate her experience of reading books. Her and others’ role-plays also showcase how human meaning-making on #BookTok is an intricate co-production of human and machine, embracing the algorithm as a co-creative agent. Critical literacy in exploratory imaginings involves recognizing this speculative synchronicity, wherein humans and algorithms seem to engage each other in real time, as creators play with and through human–algorithmic combination and integration.

Discussion and implications: building relation at the intersection of human and code

Our sample analysis of cultural production on #BookTok details how contemporary algorithmic cultures operate in processes of human–machine co-productions that are sometimes more symbiotic and at other times more antagonistic, and at all times some degree of both. This complicates notions of critical digital literacy that tend to pit user versus corporation, platform, or machine. It is not surprising that our sampling of algorithmic imaginings is shaped by many of the elements characteristic of TikTok as a whole—the predominance of humor as a strategy of engagement, memetic behavior, and meta-commentary on TikTok’s many subcultures and tropes. Also not unusual surprising is the relation-building among #BookTok community members that takes place through the shared experience of the algorithm, as users work to follow, understand, resist, explain, strategize, complain about, perform, and even celebrate the algorithm’s workings. Being literate on TikTok involves being able to mobilize the platform’s various communicative forms, which our producers do across imaginings. However, we had not anticipated that algorithmic imaginings on #BookTok would made us think differently about what critical literacy means when practiced in algorithmic cultures.

Algorithmic imaginings on #BookTok were consistently shaped by elements of critical literacy, including in its more traditional forms. Producers expressed their concern about the politics of representation—what books is the algorithm ignoring and which are constantly promoted? What authors, books, and communities of readers are marginalized or ignored? How might they use the algorithm to challenge this marginalization? They argue that corporate interests shape TikTok’s platform logics, share what they have learned about the algorithm and therefore how one might “hack” it, and in the process build a community collectively strategizing how to fight back. They also poke fun at the algorithm, which can also be an attempt at regaining a sense of control.

However, the explanatory imaginings offer elements of critical digital literacy to consider beyond the familiar. Due to the ubiquitous nature of algorithms, many users of social media platforms have reached a point of relative comfort in co-existing alongside algorithmic agents. Interestingly, users who are particularly enmeshed in very algorithmic-forward applications—such as TikTok—become comfortable enough with the algorithmic presence to try to humanize it through embodiment and other forms of role-play.

This constitutes for us another dimension of critical literacy, as producers anthropomorphize technology in a manner that addresses the symbiotic meaning-making of human and machine head-on. These producers demonstrate heightened awareness of the workings of the algorithm and platform more generally, and in their playful embodiment of FYPs, the algorithm, and the #BookTok community, seem to move beyond participation on the platform toward a re-authoring, expressing a type of control over what is largely unknown.

Regardless of the form of critical literacy offered by the various imaginings, one complication is that they all perpetuate the algorithm, even when directly critiquing, deconstructing, or attempting to subvert it. There is no “outside” of the algorithm on TikTok since every video, with its tags and other TikTok features, feeds the algorithm. This challenges the notion of critical distance from the text that has been so central to critical literacy research. It is also important to remember that the algorithms with which #BookTokers are co-creating content are the property of the platform, as is the content produced there. Algorithms are not only biased but corporate owned and controlled (e.g. ByteDance, Meta, Google). As Striphas (2015, 2023) extensively documents, in “algorithmic culture,” the work of “cultural decision-making,” including “the forms of decision making and contestation that comprise the ongoing struggle to determine the values, practices and artifacts—the culture, as it were—of specific social groups” (p. 406), has been offloaded to computers. Despite populist claims that computer algorithms merely reflect the choices of users, and therefore represent the democratization of cultural taste-making (Striphas, 2015: 407), these are proprietary, private, and linked to the production of profit through data scraping, monetizing, etc. Notions of critical algorithmic literacy or consciousness that consider the playful re-authoring of content through co-creation by TikTok producers must continue to explore the imbrication of this in the ongoing work of platform capitalism.

In a subsequent study, we are deepening our understanding of how cultural production is shaped by the algorithmic imaginary by interviewing #BookTokers themselves. We wonder how their critical awareness of the algorithm shapes not only the videos they create but also what and how they read, and how they represent their reading practices to others. This involves extending our investigation of critical literacy as a form of critical consciousness of algorithmic cultures, including algorithmically co-defined trends: How does imagining what a computational agent values influence how we perceive what others in a culture value and how we experiment and speculate, all-the-while refining the meaning we make of the books we have read? If self-representation on TikTok produces an “algorithmized self” (Bhandari and Bimo, 2022) how might it also create an “algorithmized reading self”?