Pragmatic multifunctionality of Indonesian swearing in google play app reviews: a mixed-method corpus study

1 Introduction

The proliferation of digital communication platforms has transformed the conditions under which language is produced, evaluated, and contested. Among the linguistic practices intensified by this transformation, swearing occupies a particularly unstable pragmatic position. It may attack, affiliate, entertain, intensify, resist, or complain, and these functions often emerge simultaneously rather than sequentially. Recent work in interpersonal pragmatics therefore treats swearing not merely as a lexical index of aggression but as a flexible discourse resource whose interpretation depends on target orientation, participation framework, community norms, and inferential context (Stapleton and Beers Fägersten, 2023; Dynel, 2023; Culpeper, 2021).

The present study examines Indonesian swearing in Google Play app reviews across three domains: League of Legends: Wild Rift, Netflix, and X/Twitter. Using reviews as a corpus dataset is common as shown in Prihantoro et al. (2026). The conceptual focus is pragmatic multifunctionality. By this term, the article refers to the capacity of a single swear lexeme to realize multiple, and sometimes contradictory, communicative actions depending on context (Nerlich and Clarke, 2003). This is not a claim that every swearword necessarily acquires a distinct dictionary meaning in each occurrence; rather, it is a claim about context-dependent pragmatic work, such as insulting a teammate, intensifying a service complaint, marking in-group humor, or publicly taking an ideological stance (Holmberg, 2026).

Three theoretical traditions motivate the study. The first is pragmatic research on swearing and impoliteness, which demonstrates that taboo forms can perform expressive, abusive, and idiomatic functions. Andersson and Trudgill's (1990) typology is useful here because expressive swearing releases affect without a necessary target, as in “Damn!”; abusive swearing attacks a target, as in “You stupid idiot!”; and idiomatic swearing is conventionalized, as in “I do not give a damn.” The second tradition is corpus pragmatics, which combines frequency-sensitive evidence with contextual interpretation. The third is platform-affordance theory, which explains how technological environments enable and constrain forms of social action without mechanically determining them (Landert et al., 2023; Hutchby, 2001).

The selection of League of Legends, Netflix, and X/Twitter is theoretically motivated rather than merely convenient. League of Legends represents a competitive gaming ecology in which users evaluate competence, blame teammates or systems, and release frustration through cathartic and regulatory swearing (Schweinberger and Burridge, 2025). Here, cathartic refers to the release of frustration through expletives, whereas regulatory refers to the use of targeted insults to police expected competence or community norms. Netflix represents service-oriented review discourse in which swearing often intensifies dissatisfaction with payment, registration, content availability, censorship, or application performance. X/Twitter represents social-media review discourse in which users comment on access, verification, platform changes, and public identity; swearing can therefore operate as stance-taking, mock solidarity, or ideological evaluation (Sun et al., 2021; Zhao, 2022). These platforms are comparable because all comments are Google Play reviews, yet they are contrastive because each app domain activates different communicative expectations.

Digital-platform affordances further complicate the interpretation of swearing. Anonymity can reduce accountability and intensify directness; asynchronous communication, where participants need not be online simultaneously unlike face-to-face interaction, allows users to compose affectively charged evaluations; rating systems couple textual complaint with numerical judgment; and public visibility encourages stance performance (Sha and Brookes, 2026). For this reason, the same lexeme may be pragmatically reweighted across platforms. For instance, Indonesian tai literally means feces, but in digital comments it can express disgust, as in “This game is tai,” or intensify complaint, as in “Tai, why is this app so slow?”

The study addresses four research questions. First, what categories of Indonesian swearing occur in Google Play user comments across the three platforms? Second, what pragmatic functions do these swearwords perform in context? Third, through what inferential, semantic, graphological, and platform-specific mechanisms are these functions formed? Fourth, what socio-pragmatic and cognitive contextual factors explain cross-platform variation in Indonesian digital swearing?

The practical implications are concrete. A moderation system should not assign the same pragmatic weight to aggressive tolol (stupid/idiot) targeting a teammate and solidarity-marking anjir (an emphatic variant of anjing, used here as a mild exclamation) in an X/Twitter review. App developers can use complaint-intensity patterns, such as corporeal taboo swearing in Netflix reviews, as a signal of severe service dissatisfaction rather than merely as removable toxicity. Sentiment-analysis models likewise require platform-specific lexicons because high-density swearing in League of Legends does not necessarily carry the same sentiment value as lower-density stance swearing on X/Twitter. This asymmetry follows from how each platform normalizes swearing: in League of Legends, taboo language is frequent and routinized as a channel for competence-based blame and cathartic release (Section 3.3), so any single instance carries comparatively diffuse affective weight. On X/Twitter, by contrast, users swear far less often and reserve taboo language for deliberate, marked stance-taking, so each occurrence concentrates greater evaluative and ideological force. Equal or higher swearing density therefore does not translate into equal or higher sentiment intensity across platforms.

The study also responds to persistent annotation challenges. Lexicon-only detection misses pragmatic contrasts, such as anjing (dog) as a direct insult vs. an exclamatory marker. Binary offensive/non-offensive classification fails when swearing simultaneously expresses humor, solidarity, anger, and complaint. Low inter-annotator agreement in offensive-language research further shows that context is not incidental but constitutive of interpretation (Lopez and Kübler, 2025; Leonardelli et al., 2021). The mixed-method design addresses these challenges by triangulating lexicon-based detection, KWIC interpretation, contextual coding, and statistical association tests.

2 Materials and methods

This study employs a mixed-method corpus-pragmatic design to investigate Indonesian swearing in Google Play user comments across League of Legends: Wild Rift, Netflix, and X/Twitter (Kapoor, 2022). The design integrates quantitative corpus profiling with qualitative pragmatic interpretation, enabling the study to identify not only the frequency of swear lexemes but also the communicative actions they perform. The methodological logic is abductive: distributional patterns identify candidate phenomena, while contextual analysis explains how these phenomena become meaningful within platform-specific review ecologies.

The empirical foundation comprises three Google Play user-review datasets. The original filenames are preserved in the research audit trail to support transparency and reproducibility, but the analysis refers to them as the League of Legends dataset, the Netflix dataset, and the X/Twitter dataset. Each entry consistently includes reviewId, userName, score, at (timestamp), and content. Usernames were excluded from analysis outputs and are not reproduced in this article. A comprehensive summary of the temporal coverage and data volume per platform is presented in Table 1.

Table 1. Temporal coverage and data volume of the google play user-review datasets.

The initial combined corpus comprised 500,609 raw rows. While Indonesian POS tagger is available (Prihantoro, 2025), we focus on targeted expressions, where POS tagging and lemmatization are not primary analytic units. After excluding missing content, the final analytical corpus comprised 500,599 non-empty comments. Delimiter normalization was required because the League of Legends file used semicolon delimiters with trailing empty fields. Data cleaning removed non-textual artifacts, HTML tags, and extraneous whitespace, while tokenization used whitespace and punctuation delimiters adjusted for Indonesian digital orthography. Non-standard forms were retained when pragmatically meaningful; for example, anjg, anj, anjir, and njing were treated as variants connected to the animalized-reference family but interpreted contextually rather than automatically equated.

For qualitative analysis, a stratified purposive subsample of 600 swear-containing comments was constructed, with 200 comments per platform. Within each platform, comments were stratified by rating polarity, namely negative reviews of 1–2 stars and non-negative reviews of 3–5 stars, and by the presence of high-frequency swear categories. Random selection within each stratum was conducted using Python's random sampling procedure after filtering for swear lexemes. This procedure ensured that qualitative interpretation did not rely only on the most extreme or most frequent examples.

The lexicon was compiled through four procedures: consultation of prior swearing and impoliteness literature, corpus frequency mining, manual KWIC review of candidate taboo forms, and iterative refinement during pilot coding (Seemann et al., 2023; Kharitonova et al., 2025). Initial coding of 500 comments identified recurrent forms and functions, after which categories were merged, split, or retained according to interpretive clarity and coder self-consistency. The coding scheme therefore combines theory-driven categories with corpus-driven adaptation to Indonesian digital usage. Table 2 summarizes the methodological components, illustrating the integration of corpus linguistics and pragmatic analysis within a mixed-method framework.

Table 2. Summary of methodological design components.

Digital swearing was operationalized through six non-exclusive binary categories. Each category was coded as present or absent per comment, allowing one comment to contain multiple functions. This treatment of multidimensional pragmatic categories is necessary because Indonesian swearing commonly blends insult, complaint, affective intensification, and stance in the same utterance (Bednarek and Taboada, 2025). To capture this multidimensionality systematically, Table 3 presents the six swearing categories along with their operational definitions and representative examples.

Table 3. Inter-method agreement between lexicon detector and contextual coding (Cohen's Kappa).

The moderate kappa value indicates that lexicon-based detection alone is insufficient to capture the full pragmatic range of swearing, necessitating qualitative contextual analysis for accurate interpretation (Leonardelli et al., 2021; Lopez and Kübler, 2025).

Cohen's kappa was used to measure agreement between lexicon-based detection and contextual detection (Saeed et al., 2023). The conventional interpretation used here treats values from 0.00 to 0.20 as slight, 0.21 to 0.40 as fair, 0.41 to 0.60 as moderate, 0.61 to 0.80 as substantial, and 0.81 to 1.00 as almost perfect (Landis and Koch, 1977). The chi-square test assessed the association between swearing, coded as present or absent by lexicon detection, and rating category, coded as negative for 1–2 stars and non-negative for 3–5 stars. Odds ratios were interpreted as evidence of association between swearing and negative ratings, not as standalone validation of the detector. Because multiple platform-level comparisons were performed, results are interpreted conservatively; the corpus size makes small effects statistically detectable, so substantive interpretation prioritizes effect direction, platform contrast, and qualitative triangulation.

KR-20/Cronbach alpha was used cautiously. A low value does not necessarily indicate instrument failure because the six categories were designed to capture a multidimensional construct rather than a unidimensional psychological scale. Split-half reliability assessed whether platform-level swear rates remained stable across deterministic corpus halves. Spearman's rho was used because ratings are ordinal and swear counts are non-normally distributed with many zeros; nevertheless, the results are treated as exploratory rather than definitive.

A replication protocol was maintained through documented preprocessing parameters, an auditable lexicon table, and AntConc settings. KWIC analysis used AntConc 4.3.1 with manual inspection of immediate concordance lines and expanded context where required. The anonymized corpus and analysis scripts are available from the corresponding author upon reasonable request, subject to Google Play terms of service and privacy restrictions. A machine-readable version of the lexicon can be produced from Appendix A.

Ethically, the study uses publicly accessible app-review data, anonymizes personal identifiers, and avoids reproducing usernames. The purpose of analysis is linguistic and socio-pragmatic rather than evaluative of individual users. Offensive examples are included only where analytically necessary and are translated to support international readership.

3 Results

3.1 Temporal coverage of the corpus

The temporal span of the corpus differs by platform but is sufficient for comparative corpus-pragmatic analysis. League of Legends comments cover October 27, 2020, to August 6, 2024; Netflix comments cover December 6, 2013, to July 10, 2024; and X/Twitter comments cover January 28, 2018, to July 31, 2024. All datasets contain parseable date metadata for virtually every row, enabling temporal contextualization even though the present article does not claim to provide a full diachronic model of swearing change.

3.2 Validity and reliability of swearing detection instruments

Table 4 presents the validity and reliability evidence. The overall Cohen's kappa is 0.3858, which is best interpreted as fair agreement. This result is analytically important because it demonstrates that lexicon-based detection and contextual pragmatic detection overlap but are not interchangeable. League of Legends reaches moderate agreement, whereas Netflix and X/Twitter remain in the fair range, indicating that service complaints and social-media stance-taking create more ambiguous pragmatic environments for swearing interpretation.

Table 4. Corpus profile by platform: review counts, ratings, token statistics, swearing prevalence, and emphasis markers.

The fair kappa value does not weaken the study; rather, it clarifies the object of inquiry. If lexical swearing were pragmatically transparent, lexicon-based and contextual detectors would converge strongly. Their partial divergence instead supports the central claim that Indonesian digital swearing is context-sensitive, multifunctional, and resistant to binary classification.

3.3 Corpus profile and score distribution

The combined corpus contains 500,609 raw reviews and 500,599 non-empty comments after cleaning. Netflix has the lowest mean rating and the highest negative-rating percentage, while League of Legends and X/Twitter show more polarized distributions with many 1-star and 5-star reviews. Swear-review percentages are similar for League of Legends and Netflix, but X/Twitter has a substantially lower percentage, suggesting that social-media reviews may use swearing more selectively for stance and affect rather than as a frequent complaint marker (Schweinberger and Burridge, 2025).

These cross-platform differences in rating patterns and swearing prevalence are summarized in Table 5, which presents the full corpus profile including review counts, token statistics, swearing density, and emphasis marker usage.

Table 5. Distribution of user ratings by platform (counts and percentages).

Taken together, Tables 5, 6 reveal that Netflix exhibits the highest proportion of negative ratings (53.7%) alongside a moderate swearing prevalence (5.7%), suggesting that user dissatisfaction on app review platforms often coincides with swear use (Lewandowska-Tomaszczyk et al., 2023). League of Legends and X/Twitter, by contrast, show more polarized rating distributions—dominated by both 1-star and 5-star reviews—yet diverge sharply in swearing frequency: LoL users swear nearly twice as often as X/Twitter users despite similar rating patterns. This indicates that swearing is not merely a function of negative sentiment but is shaped by platform-specific communicative norms, with gaming reviews favoring expressive intensity and social media reviews employing swearing more sparingly for targeted stance-taking.

Table 6. Distribution of swearing categories by platform (number and percentage of reviews).

3.4 Swearing categories and lexical frequencies

The six-category classification reveals cross-platform regularities and differences. Cognitive devaluation and animalized references are prominent across platforms, but their pragmatic load differs: in League of Legends, they often evaluate competence; in Netflix, they intensify complaint; and in X/Twitter, they may become stance markers or ironic exclamations (Fortuna et al., 2021). Corporeal taboo is proportionally highest in Netflix reviews, where users frequently compress intense service dissatisfaction into short, affect-heavy comments. The cross-platform regularities and differences outlined above are systematically captured in Table 7, which quantifies the distribution of each pragmatic category alongside representative swear words per platform.

Table 7. Top swear lexemes by platform with frequency counts.

While Table 7 reveals proportional patterns across platforms, Table 8 further refines the analysis by identifying the most frequent swear lexemes per platform with their English translations, thereby bridging category-level trends and lexical-level usage.

Table 8. Validity and reliability evidence for swearing detection instruments.

Building on the lexical and distributional evidence from Tables 7, 8, the findings align with prior studies that situate gaming and evaluative online environments as intensifiers of impoliteness, while app reviews convert user dissatisfaction into condensed affective judgments (Culpeper, 2021; Zhang et al., 2025; Hadi and Fard, 2023). The Indonesian data refine this generalization by showing that the same lexical families are pragmatically redistributed according to app domain. Figure 1 visualizes the distribution of the six swearing categories across the three platforms and makes visible the category-platform interaction that is less immediately apparent in raw frequency tables.

The heatmap shows that cognitive devaluation and animalized reference dominate League of Legends, while Netflix displays a comparatively stronger corporeal-taboo profile. X/Twitter has lower percentages overall but maintains visible animalized-reference use. The figure therefore supports the argument that pragmatic functions are platform-conditioned rather than uniformly distributed. Figure 2 compares swear-review percentage with swear-token density per 1,000 tokens in order to distinguish prevalence from intensity.

The comparison shows that Netflix and League of Legends have similar swear-review percentages, yet Netflix has the highest density. This means that Netflix swearing is more concentrated in shorter review texts, which is consistent with service-complaint intensification. X/Twitter has a lower swear-review rate, but its swear density indicates that swearing remains pragmatically salient when it occurs.

3.5 Qualitative KWIC analysis with translations

The KWIC analysis demonstrates that swearing functions as a situated pragmatic resource rather than as a mechanically offensive token (Roy and Biswas, 2024). Translations are provided to make the Indonesian data accessible to international readers. The examples are reproduced as user-generated digital discourse; non-standard spelling is therefore retained where it contributes to pragmatic force. To illustrate how these pragmatic functions manifest across platforms and contexts, Table 9 provides representative KWIC examples with glosses, full English translations, and interpretive commentary.

Table 9. KWIC examples, translations, and pragmatic interpretations.

Across platforms, three patterns recur. First, swearwords frequently index evaluative stance, but the object of evaluation differs: co-player competence in League of Legends, service failure in Netflix, and access or platform-change frustration in X/Twitter. Second, graphological features such as capitalization, repeated punctuation, and vowel or consonant lengthening operate as pragmatic amplifiers. Third, exceptions are analytically important: anjir and anj may become exclamatory or humorous rather than abusive, which explains why contextual coding diverges from lexicon-based detection.

4 Discussion

4.1 Core findings: types and functions

The results show that Indonesian swearing in Google Play reviews is best understood as pragmatic multifunctionality. The dominant categories, cognitive devaluation and animalized reference, are not merely lexical classes; they are resources for evaluating competence, dramatizing dissatisfaction, negotiating solidarity, and performing public stance. League of Legends foregrounds competence-based blame because the review ecology is tied to competitive play. Netflix foregrounds complaint intensity because users evaluate paid service, content availability, and app performance. X/Twitter foregrounds stance and identity because the reviewed application itself is a social-media infrastructure associated with public discourse.

This finding refines the aggression-centered view of swearing. Swearing is often associated with negative ratings, as the odds ratios show, but the association does not exhaust its pragmatic work. A swearword may be negative in affect while still affiliative in function, as when users share frustration through humorous exclamation. Conversely, a lexeme may appear mild in form but aggressive in context if it targets a user, group, or platform actor directly.

4.2 Mechanisms: how pragmatic multifunctionality emerges

Pragmatic multifunctionality emerges through inferential modulation. A lexeme carries a literal or historically recoverable meaning, but its situated function depends on target, co-text, platform genre, graphology, and shared community expectations. The mechanism can be described as a movement from lexical form to contextual cueing and then to pragmatic uptake (Retta, 2023). Thus, anjing does not automatically mean the same thing in every review; its pragmatic value is constructed through the interpretive relation between word, target, and platform setting.

These cross-platform differences in how anjing shifts from literal meaning to pragmatic function are systematically captured in Table 10, which compares the lexeme across League of Legends, Netflix, and X/Twitter.

Table 10. Inference mechanisms for the lexeme anjing across platforms.

The inferential pathways illustrated in Table 10 show that pragmatic function emerges from the interplay of lexical meaning, contextual cues, and platform norms—a layered process that Figure 3 visualizes as a structured model (Yuliawati et al., 2024; Surahmat et al., 2024). The lexical-semantic layer supplies taboo potential, but pragmatic function is produced only when lexical meaning interacts with platform affordance and socio-cognitive inference (Chae, 2026; Tomlinson, 2026; Rodríguez et al., 2025; Wang et al., 2026; Jumle, 2026; Babatunde et al., 2026). Figure 3 illustrates how interpretation involves lexical-semantic, pragmatic-functional, platform-affordance, and socio-cognitive layers.

The model demonstrates that no single layer determines meaning. A swearword becomes an insult, complaint, joke, or solidarity marker only through the interaction of literal association, speaker stance, platform context, and community inference. This layered explanation directly addresses why the same lexeme can produce different pragmatic effects across the corpus.

Graphological strategies further contribute to meaning formation. Capitalization, repeated punctuation, and lengthening such as anjingg or anjirrr function as written substitutes for prosody, intensity, and affect. Code-mixing also expands pragmatic range, especially in forms such as fak, which borrow the affective force of English profanity while being localized through Indonesian orthography.

4.3 Platform affordances as explanatory framework

Platform affordances explain why the same lexicon is unevenly distributed and differently interpreted. The relevant affordances include anonymity, public visibility, asynchronous review posting, star ratings, app-specific complaint norms, and the absence of immediate conversational repair. These affordances do not determine meaning; rather, they condition the kinds of pragmatic action that become salient and recognizable.

Figure 4 maps the three app domains onto their dominant communicative ecologies within the shared environment of Google Play reviews.

League of Legends is associated with competitive flaming, competence evaluation, and catharsis (Hadi and Fard, 2023). Netflix is associated with consumer complaint, service dissatisfaction, and ideological critique. X/Twitter is associated with public stance, identity positioning, and ironic commentary. The figure therefore clarifies why the study compares these platforms: they share a review infrastructure while differing in the social actions' users perform through reviews.

Participatory cultures also differ. League of Legends reviews import the interactional logic of competitive play, where blame, expertise, and rank are central. Netflix reviews import the logic of consumer entitlement and service accountability. X/Twitter reviews import a broader social-media repertoire, including public critique, humor, identity display, and collective reaction to platform change (Holmberg, 2026; Sengul, 2025; Zarate et al., 2023). These differences operate at multiple discourse levels: word choice, clause structure, graphological emphasis, interactional sequence, and platform-wide norms.

4.4 Methodological contributions and limitations

Methodologically, the study demonstrates the value of combining large-scale corpus evidence with contextual pragmatic interpretation (Altamimi, 2026). The corpus size supports robust descriptive profiling, while KWIC analysis prevents lexical frequency from being mistaken for pragmatic meaning. The fair overall kappa value is especially instructive because it empirically identifies the boundary of lexicon-only detection. A detector may identify a token, but it cannot reliably determine whether that token attacks, jokes, complains, affiliates, or intensifies without context. Figure 5 summarizes the inference chain from data ingestion to integrated interpretation.

The pipeline highlights that computational detection is only one stage in the analysis (Liu et al., 2023). Cleaning, lexicon matching, KWIC extraction, contextual coding, reliability assessment, and theoretical interpretation must be treated as mutually checking procedures. This structure also improves replicability because each decision point can be audited.

The limitations are equally important. First, the overall inter-method agreement is fair, not moderate, which confirms the difficulty of annotating pragmatically nuanced swearing. Second, the study relies on a single primary annotator, although self-consistency checks, coding documentation, and KWIC verification mitigate but do not eliminate subjectivity. Third, Google Play comments represent a specific digital genre and should not be generalized uncritically to chat, in-game communication, comment threads, or private messaging. Fourth, the temporal coverage ends in 2024, and platform norms may have shifted since then. Fifth, the statistical analyses follow corpus-linguistic conventions but have not been independently reviewed by a statistician; future work should examine multiple-comparison corrections, alternatives for count data with many ties, and additional robustness checks.

4.5 Theoretical and practical implications

Theoretically, the study contributes to corpus pragmatics by showing how taboo forms become socially meaningful through the interaction of lexical potential, platform affordance, and socio-cognitive inference (Çiftçi, 2025; Ramli et al., 2026; Blake et al., 2026; Palmer, 2026). It also contributes to multilingual offensive-language research by foregrounding Indonesian, a language whose digital swearing practices are shaped by regional forms, orthographic variation, code-mixing, and culturally specific insult systems (Ranasinghe and Zampieri, 2022). Figure 6 places the observed functions on a continuum from low-attack solidarity and intensification to high-attack targeted abuse.

The continuum avoids a binary offensive/non-offensive model. It shows that swearing can move gradually from exclamation to complaint, ideological stance, and targeted abuse depending on contextual dependence and attack orientation (Zhang et al., 2025; Cicek et al., 2026). This is particularly relevant for moderation because the same token may fall at different points on the continuum in different contexts.

Practically, the findings suggest that Indonesian content moderation should incorporate platform-sensitive and function-sensitive annotation (Nascimento et al., 2023; Qureshi et al., 2026). A word list can flag candidate comments, but moderation decisions should consider target, stance, rating, graphological intensity, and platform domain (Moreno-Ortiz and García-Gámez, 2025). Sentiment-analysis systems for app reviews should likewise avoid transferring weights from one platform to another: high swear density in gaming reviews may index normative frustration, while similar density in service reviews may indicate severe consumer dissatisfaction. For app developers, swearing patterns can function as diagnostic signals of friction points, especially when clustered around payment, verification, matchmaking, or content access.

5 Conclusion

This study provides preliminary but substantial evidence that Indonesian swearing in Google Play reviews is pragmatically multifunctional. Across 500,599 non-empty comments, swearwords function not only as aggression but also as complaint, catharsis, humor, solidarity, evaluative intensification, and ideological stance-taking. Platform comparison shows that these functions are patterned: League of Legends foregrounds competence-based blame, Netflix foregrounds service dissatisfaction, and X/Twitter foregrounds public stance and identity positioning.

The study's central contribution is not the claim that swearwords always possess multiple dictionary meanings, but that their pragmatic force is contextually negotiated. The fair overall kappa between lexicon-based and contextual detection confirms that lexical identification is necessary but insufficient. Consequently, future Indonesian offensive-language research should combine lexicon resources with contextual annotation, graphological analysis, and platform-sensitive interpretation. Several limitations warrant caution (Ljubešić et al., 2023). The single-annotator design constrains generalizability, the data are restricted to Google Play review discourse, temporal coverage ends in 2024, and the statistical procedures require further robustness testing and independent statistical review. Even with these limitations, the article offers a grounded model for studying Indonesian digital swearing as a culturally situated, platform-conditioned, and inferentially complex discourse practice.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Generative AI tools were used solely for language editing, grammatical refinement, improving readability, and aligning the manuscript with the journal format. No AI tool was used to generate the research design, collect or analyze data, interpret results, or formulate conclusions. All outputs were critically reviewed, verified, and approved by the author(s), who take full responsibility for the content, accuracy, and integrity of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: John T. Jost, New York University, United States

Reviewed by: Sebastian Sadowski, Radboud University, Netherlands

Claria Dewa Ayu Kadek, Universitas Warmadewa, Indonesia