Daejeon, July 27–31
This study investigates how diverse online communities interpret and identify with KPop Demon Hunters (2025), a Korean American animated feature film released on Netflix. The film pairs Korean-inflected pop aesthetics and mythic references with a familiar coming-of-age grammar organized around concealment, shame, and eventual self-acceptance. This layered structure helps explain the film’s strikingly heterogeneous reception across Reddit, Letterboxd, and YouTube, where viewers mobilize mixed feelings including joy, irritation, pride, nostalgia, and ambivalence while disputing representation, genre expectations, fandom norms, and cultural ownership. Rather than treating online discourse as a source of measurable sentiment or a corpus whose primary value lies in large-scale pattern detection, this project approaches reception as a constellation of historically and culturally situated interpretations. It asks two linked questions: How do viewers from different cultural and ideological positions claim the same story as their own across platforms with different participation norms? What emotional and narrative vocabularies organize these claims into recurring, but contested, interpretive formations?
The project intervenes in a methodological tension within digital humanities and reception studies. Large-scale approaches including sentiment analysis, topic modeling, and network analysis can trace patterns of attention and discourse at scale (Moretti 2013; Jockers 2013). Yet critical data studies caution that big-data claims of neutrality obscure how sampling, classification, and platform infrastructures shape what can be seen and said (boyd / Crawford 2012). Arguments that “raw data” do not exist outside historical and institutional production invite scholars to treat digital traces as socially produced artifacts rather than neutral inputs (Gitelman 2013). Data feminism underscores that data practices are structured by power, and that methodological choices determine whose experiences become legible and whose remain erased (D'Ignazio / Klein 2020). These critiques suggest that purely large-scale approaches risk converting interpretive complexity into an average attitude or abstract network structure, analytically inadequate when the object of study is plural, contested meaning-making.
Fan and reception studies traditions offer essential resources because they treat reception as socially organized meaning-making rather than as private reaction. Audience reception research describes this meaning-making as coordinated through interpretive communities, where shared premises of interpretation circulate, stabilize, and become sites of dispute (Lindlof 1988). Fan studies, in turn, shows how such communities establish norms for what counts as legitimate knowledge and proper attachment, and how they police those boundaries through practices of belonging and gatekeeping (Jenkins 1992; Gray et al. 2007). English and literary-cultural studies likewise frame reading as historically situated interpretation, emphasizing that meanings emerge through socially learned repertoires rather than residing transparently in the text (Iser 1978; Jauss 1982; Fish 1980). Building on these lineages, this project uses an explicitly data-modeled workflow to render those premises and repertoires visible and citable: coding and memoing formalize how platform-specific norms and community practices organize recurring, contested interpretive formations without collapsing them into aggregate sentiment. Within DH, this approach aligns with arguments that modeling and structured annotation are central forms of knowledge production rather than mere preprocessing (Flanders / Jannidis 2016).
To move beyond the limitations of large-scale approaches, the study foregrounds interpretive small data: a curated corpus of approximately one hundred publicly accessible posts and comments drawn from Reddit, Letterboxd, and YouTube. These platforms support distinct reception genres: threaded argumentation, review and rating cultures blending fan and critic registers, and short affectively charged responses to clips and trailers. The corpus spans the film's initial release (June 2025) through later waves of festival and award circulation. Sampling is purposive rather than representative, designed for maximum variation in stance, affect, and interpretive focus, including enthusiastic praise, sharp critique, and ambivalent responses. This design aligns with small data, thick data approaches that slow down trace-based research to situate posts within community practices and platform norms (Latzko-Toth et al. 2016). It also draws on digital ethnography’s insistence that online discourse is “embedded, embodied, and everyday,” shaped by infrastructures and by the entanglement of online practices with offline identities (Hine 2015).
The study implements a three-stage coding and memoing workflow adapted for digital-humanities data modeling. Stage 1 applies descriptive coding across the corpus, recording contextual fields (platform, date, post type) and thematic and affective labels including representation, gender, and culture alongside markers such as joy, anger, disappointment, and relief. Stage 2 conducts interpretive coding on especially rich or contested posts, shifting attention from what a post is about to what it is doing: how viewers perform recognition, critique, ambivalence, or self-narration. Analytic memos capture emergent tensions, enabling iterative movement from individual posts toward synthetic claims without collapsing difference into a single metric (Saldaña 2021). Stage 3 identifies interpretive constellations: recurrent combinations of themes, affects, and interpretive moves derived from coded co-occurrence and refined through memo-supported close reading, treating the codebook and memos as reusable interpretive infrastructure.
Unlike distant reading, which abstracts patterns across large corpora, this project uses computational annotation not to scale up but to scale down with precision, rendering interpretive decisions traceable and the analyst's positionality visible. To theorize how constellations cohere without presuming stable demographic categories, the project draws on Papacharissi's concept of affective publics, where networked discourse gathers through connective storytelling and shared feeling rather than ideological consensus (Papacharissi 2015). This lens explains why the same textual elements can be recruited into incompatible narratives: one constellation may read concealment as a queer coming-of-age script while another treats the same arc as moral reassurance; one may mobilize the film as Asian-diasporic recognition while another disputes authenticity through fandom gatekeeping. The analytic aim is not to decide which reading is correct, but to model how multiple meanings become socially actionable through platformed affect and narrative repertoires.
The study treats data politics, privacy, and reproducibility as integral to method, adopting cautious ethical practices including pseudonymization and selective quotation for sensitive disclosures, and pursuing reproducibility through procedural transparency rather than full corpus release, sharing sampling logic, codebook structure, and memo conventions as reusable methodological artifacts. The study also acknowledges that purposive sampling, while designed for maximum variation in interpretive stance, cannot guarantee demographic representativeness across axes such as gender, race, or geographic location, a limitation future work might address through expanded corpus design. Substantively, the paper produces a fine-grained map of plural reception that resists reduction to average sentiment or dominant-topic summaries. Methodologically, it proposes interpretive constellation modeling as a replicable digital-humanities approach to reception studies, one that preserves heterogeneity without flattening it into aggregate metrics and renders annotation itself a form of critical data practice.