Netflix allegedly runs some 36,000 categories within its recommendation algorithms. The home screen—the first page after login—displays unique arrays of micro-genres such as Witty Mockumentaries, which includes dry, satirical “documentaries” like The Office, and Reluctant Adults, stories about characters struggling with “grown-up” responsibilities, featuring shows like Unbreakable Kimmy Schmidt and Beef. This marks a significant departure from traditional genre categories such as action & adventure or romantic comedy. Such fine-grained fragmentation emerges from the need to pinpoint viewers’ interests and maximize the efficiency of recommendation systems—what might be called a form of personalized programming.
This system reduces the possibility of choice overload by filtering out titles a viewer is unlikely to enjoy. At the same time, it risks placing the viewer within a loop of likable content, limiting the chance for unexpected discovery. By tracking what and when a viewer watches, Netflix assigns users to one of roughly 2,000 “taste communities”: Supernatural Romance (fans of Twilight and The Vampire Diaries), Global Health-Conscious Foodies, or The Anti-Hero. These communities function as more than a classificatory system. They begin to approximate the role of a cultural critic, mapping media experiences into clusters of meaning that appear specific to cultures, regions, or viewing habits.
Is Netflix, then, not only occupying viewers’ time but also reconfiguring how media content is produced and understood? The Glory offers a kind of litmus test for this question.
Written by Kim Eun-sook, known for Guardian: The Lonely and Great God and Mr. Sunshine, The Glory is a revenge thriller in which Dong-eun, a victim of brutal school bullying, spends years constructing a meticulous and calculated revenge against her perpetrators and the bystanders who enabled them. Within Netflix’s micro-genre system, the series might be categorized under labels such as Dark Korean Revenge, Psychological Thriller, or Social Issue Drama. At first glance, these appear to be neutral descriptors of plot and tone. Yet, when combined with viewer behavior, they become the foundation of taste communities.
Viewing duration, sequence (what is watched before and after), time of day, and search history are all measured to construct these clusters. Metrics such as completion rate, binge speed, and rewatch value are weighted to infer patterns of desire. What emerges is not simply behavioral data, but an interpretive layer: what these patterns are taken to mean about the viewer.
Taste communities thus function as both market segmentation and cultural definition. They determine not only who the viewer is, but what the content does for them. For instance, a viewer who completes The Glory and then watches documentaries on the Menendez brothers may be grouped into a “justice-driven revenge” cluster—viewers drawn to narratives of calculated retribution against corrupt or powerful figures. Alternatively, they may fall into a “dark social realism” cluster, gravitating toward narratives that engage with systemic violence, class disparity, or trauma.
These clusters, in turn, shape future recommendations. More significantly, they inform future production. The algorithm does not merely predict taste; it operationalizes it. This is where its power lies—not only in refining recommendation systems, but in defining what content means for viewers, and what kinds of stories are worth telling next.
Before the Code
School bullying and violence constitute a serious social issue. Around the early 2000s, the press and media began to address more systematically both the short- and long-term effects of this pervasive problem. Victims often suffer lasting psychological damage, ranging from anxiety and depression to Post-traumatic stress disorder. These effects extend beyond the individual, undermining trust within communities and schools. The consequences are not limited to victims; research suggests that aggressors are also more likely to engage in criminal behavior later in life. What emerged was a troubling recognition: children and adolescents could be exposed to deeply traumatic experiences within institutions meant to protect them. In response, institutional efforts such as the School Violence Countermeasures Review Committee have sought to address both prevention and intervention, though the problem persists.
Since around 2020, a notable wave of Korean drama series has taken up the theme of school violence, often reframing it through genre conventions rather than social realism. Weak Hero Class 1 follows Yeon Si-eun, a top-ranking student who appears physically weak but strategically resists violence through intellect and psychological insight. The King of Pigs, adapted from an acclaimed animated film, presents a darker trajectory: a man traumatized by school bullying becomes a serial killer, leaving cryptic messages for a former friend turned detective. Meanwhile, All of Us Are Dead adopts the structure of a zombie apocalypse, grounding its outbreak in a history of sustained bullying and institutional failure. Across these examples, school violence is no longer presented solely as a social issue but is mediated through the logics of psychological thriller, revenge narrative, and horror.
At one level, the emergence of such genre-inflected narratives may appear coincidental with the rise of streaming platforms such as Netflix, with their micro-genre categorizations and taste communities. Yet it is difficult to overlook the extent to which these classificatory systems exert influence on production. In an environment where visibility is algorithmically mediated, creators are increasingly compelled to situate their work within recognizable grids of taste and interest. Without the prior existence—or at least the algorithmic anticipation—of a viewership for a category such as “dark revenge thriller,” a series like The Glory would be far less likely to emerge.
What appears, then, as a simple act of categorization begins to function as a structuring logic. It not only organizes existing content but also shapes how social issues are narrativized in the first place. In this sense, algorithmic systems do travel across cultures as neutral technologies; they carry with them a set of narrative probabilities—formats, tonalities, and expectations—that can be readily adapted to local contexts. School violence, as a culturally specific and socially urgent issue, becomes legible within globally circulating genre frameworks. The result is a form of cultural translation in which local trauma is rearticulated through universally recognizable narrative codes.
Perform the Reading
Algorithms are often understood as AI-driven systems that take inputs—viewing patterns, clicks, durations—and produce outputs in the form of recommendations. Yet they are not purely technical. They operate at the intersection of non-human computation and human culture, where data is not only processed but interpreted. What emerges is not simply a system of calculation, but a dynamic field in which viewers and algorithms continuously shape one another.
Viewers, in this sense, are not passive recipients of algorithmic logic. They actively engage with these systems, finding ways to interact with, adjust, and at times manipulate the patterns that define them. What becomes significant, then, is not only how algorithms function, but what viewers do with them. This is where a mode of engagement that may be described as performative reading begins to take shape.
Performative reading unfolds through concrete practices. Viewers “clean” their profiles by removing titles from watch histories, preventing so-called guilty pleasures from shaping future recommendations. They deliberately “like” or binge-watch certain genres to align themselves with preferred categories, even when these do not reflect their immediate interests. Some create multiple profiles—not for different individuals, but for different “mood personas”—switching, for instance, between a documentary-oriented profile and one dedicated to light entertainment. In doing so, they do not simply consume content; they performatively manage the algorithm’s construction of their “taste.”
Such behaviors do not exist outside the algorithmic system. They unfold within it, introducing small but significant deviations. The viewer does not escape classification but complicates it, producing patterns that are only partially legible within existing taste communities. In this sense, performative reading operates as a form of friction—an ongoing negotiation between prediction and experience.
This dynamic also feeds back into the system itself. A series such as The Glory may initially be positioned within a category such as “dark revenge thriller.” Yet its classification is not fixed. As viewer behaviors shift—through mismatched signals, unexpected viewing patterns, or social reinterpretations—the meaning attributed to the series may also change. Viewers may respond less to its revenge structure than to its melodramatic affect, its performances, or its emotional intensity. What appears as a stable category begins to loosen under the pressure of lived engagement.
This does not suggest that viewers are fully free from algorithmic systems. Media production and consumption remain deeply intertwined with technological infrastructures that shape both visibility and meaning. Rather, it points to a more complex condition in which the viewer occupies a shifting position between two modes of subjectivity: the algorithmic agent—measured, predicted, and categorized—and an update subject capable of reorienting and reinterpreting those classifications.
Performative reading marks the space in which this negotiation takes place. It is not a rejection of the algorithm, but a way of inhabiting it differently. Within a system designed to anticipate preference and stabilize meaning, the act of viewing retains a limited but persistent capacity to deviate, to reconfigure, and to exceed the coordinates assigned to it.