This study examines the evolution of gender power diferences within the Omegaverse trope on the Archive of Our Own (AO3) fanfiction platform. The Omegaverse, a fan created trope defined by distinct secondary gender categories (Alpha, Beta, Omega), provides a lens to explore and critique gender dynamics. We used connotation frames and the Riveter NLP pipeline to measure power dynamics in Omegaverse slash relationships between characters. Our corpus includes English stories across eight prominent fandoms, including Japanese anime, K-pop, Western TV series, and Western films. Our findings show Alphas generally exhibit higher power scores than Omegas, consistent with the trope's constructs, but with notable variance between fandoms, while most fandoms exhibit more within-group consensus when more fans start writing. This research ofers insights into how fan communities evolve and challenge traditional power structures.
Across a variety of cultural fields, researchers have identified pervasive gender inequality, characterized by the underrepresentation and marginalization of women. This disparity is evident both in who is able to produce cultural works and who is depicted within them12[]. Gender, conceptualized as a societal construct, has historically played a pivotal role in shaping the narratives and themes of literary works. In contemporary literary scholarship, the examination of gender biases and representations emerged as a critical area of study1][.
Within literary genres, fanfiction stands out as a highly popular storytelling form,
extensively accumulating a large number of stories on online platforms fueled by the development
of digital social reading 1[
One distinctive feature of fanfiction is using original characters or settings, derived from
TV shows, books, movies, or other media to create new stories 1[
The Omegaverse, or Alpha/Beta/Omega (ABO) Dynamics, ofers an inventive approach to
tackling gender stereotyping and associated prejudices8[]. Inspired by ”wolf pack hierarchies”,
this fictional framework introduces a speculative social system where characters are divided
into three biologically-driven categories: alpha, beta, and omega, to intensify social and sexual
dynamics [
The concept of secondary genders such as Alpha, Beta, and Omega allows characters of the
primary gender, both female or male, to become leaders like Alphas, diminishing the opposition
between the primary genders of men and women to some extent. Nonetheless, the design
of this worldview with secondary genders also emphasizes that biology ostensibly dictates
sexual and social roles, and it introduces gender hierarchy into queer fanfiction more freely and
exaggeratedly [18][19][
Yet, scholars also note that the gender hierarchies and notions depicted in Omegaverse
fanfiction narratives contrast with the foundational worldviews of the Omegaverse setting. Popova
[18] argues that rather than uncritically adopting gendered sexual scripts and socially
constructed power diferentials, Omegaverse stories are employed by their creators and audiences
as a critical tool. These narratives problematize power and consent, ofering avenues for
negotiating meaningful, consensual intimate relationships within broader abusive social structures.
Goldmann [
If there exists a characteristic of reflection and correction within the Omegaverse trope, then the formation of this characteristic should also be the result of continuous evolution and iteration. Yet current research focused on the gender hierarchy represented in the Omegaverse has not started from the development trajectory of Omegaverse fanfiction, to comb through whether there are dynamic changes and overall trends in the portrayal of gender hierarchy as it rapidly develops and attracts an increasing audience.
Therefore, the focus of this study is on the gender-based hierarchies of social and sexual power in the Omegaverse trope and how these hierarchies evolve and interact with the trope’s increasing popularity. The primary research question is: Are there gender-based power diferences that are frequently occurring along with the Omegaverse trope?
To explore this question, we address two subquestions: (1) How do the power dynamics of Alpha, Beta, and Omega characters in slash relationships difer across various fandoms? Is there a significant diference between top fandoms? (2) Do these dynamics evolve over time?
We utilize the GOLEM database [
As illustrated in Figure 1A, there has been a notable upward trend in the publication of Omegaverse trope fanfiction since 2010. Considering the simultaneous expansion of the AO3 platform, we utilized the annual total number of published fanfiction on AO3 to reveal the growth rate of the ABO trope relative to the overall fanfiction production (Figure 1B). What stands out is not only the increasing number of ABO stories but also their expanding proportion within the entire fanfiction corpus over time. This indicates a growing interest in the ABO theme among AO3 users. 1Due to copyright restrictions, the story content cannot be made publicly available . The corpus metadata, our code, and computation results can be accessed athttps://github.com/GOLEM-lab/Omegaverse_Study/
M/M pairings, which depict romantic relationships between male characters, constitute a significant portion of AO3, accounting for 46% of the total. This trend is even more pronounced within the Omegaverse, where such stories represent 70% of the total. Furthermore, 4.23% of all fandoms on AO3 include at least one Omegaverse fanfiction. Fandoms dominated by male leading characters, such as the Japanese Shounen animeMy Hero Academia, the Korean pop boy band BTS, and Marvel universe, are more prevalent in Omegaverse fanfiction compared to the general AO3 trend.
Following this, based on observations of popular fandoms and corresponding slash relationships within the Omegaverse corpus, we selected eight slash relationships from the top 15 in the original Omegaverse corpus to build a narrower corpus. These relationships originate from Japanese anime, K-pop, Western TV series, and Western film universes. These eight slash relationships account for 12.8% of all ABO stories in the AO3 platform and span the timeline from 2010 to 2022, representing diferent stages of popularity. This selection balances Eastern and Western cultural industries, as well as diferent types of media, ensuring a relatively robust sample of the overall creation trends within the Omegaverse trope.
Within the domain of computational linguistics, exploring gender dynamics typically involves delving into the descriptive language surrounding persons, focusing particularly on the behaviors/verbs or nouns/adjectives employed in their representation14[].
We first analyze the proportions of verbs and adjectives in the corpus, with results showing a higher proportion of verbs. Consequently, for this initial analysis, we choose to focus on verbs to measure gender power diferences. Then we leverage information from online fan encyclopedia to enhance the accuracy of character coreference resolution. Next, we identify the secondary genders of characters and calculate the gender power diferences between Alpha and Omega characters in slash relationships. Finally, we critically analyze the computational result together with close reading and human evaluation.
Using SpaCy2, we conducted a statistical analysis of the parts-of-speech (POS) distribution within the Omegaverse corpus. We compared this distribution to a benchmark of CMU Movie Summary Corpus 3[] since connotation framework was set for studying gender dynamic in movie summary scripts (Table 1). Our analysis revealed that both corpora have a high proportion of verbs, though the Omegaverse corpus exhibiting a slightly lower frequency of verbs, adjectives, and adverbs, and a higher frequency of pronouns compared to the CMU Movie Summary Corpus. The shared high proportion of verb usage makes verb-based gender power measurement methods suitable for analyzing online fanfiction.
We initially employ the Python library Riveter2][ and its power-agency verb lexicon to quantify gender hierarchy within fanfiction texts. Subsequently, we use close reading methods to verify the quality of the measurement results. Riveter is a natural language processing (NLP) pipeline designed to analyze the dynamics of power and agency between entities in texts, including characters, thereby testing whether fanfiction’s transformative potential can be substantiated with quantitative methods.
To quantify the levels of agency and power associated with each character in the texts,
Riveter leverages the theoretical foundation of Connotation Frames and the power and agency
lexicons developed by Sap et al. 2[
Sap et al. [21] enriched this framework with a lexicon of verbs, each assigned scores based
on their usage frequency in relevant corpora, such aNsew York Times articles and movie script
corpora, for sentiment and power and agency frames respectively. This evolving interest in
Connotation Frames among researchers, especially within cultural analytics and digital
humanities, shows their utility in exploring the framing of interpersonal dynamics across diverse
datasets and inquiries [
To identify gender hierarchy in literature, it’s crucial to accurately recognize the gender of ifctional characters within the text. However, in Omegaverse fanfiction, our focus on gender hierarchy revolves around the secondary genders of Alpha, Beta, and Omega. Additionally, since M/M slash relationships dominate the Omegaverse trope, with male characters assigned diferent secondary genders, traditional pronoun and name-based gender classification methods are inefective.
Instead, we organize characters based on explicit secondary gender attributes provided by authors in metadata tags. Using this information, we could detect the secondary gender attributes of corresponding characters. Then, we employ the Riveter pipeline to perform named entity recognition and coreference resolution to calculate social dynamic scores for the two relevant characters involved in slash relationships in fanfiction. To enhance recognition comprehensiveness, we utilize character names and aliases retrieved from the wiki website fandom.net, as well as character-related names and nicknames appearing in fanfiction metadata, as the basis for named entity recognition. After running the Riveter pipeline to get the power score per character, we delete the stories with extreme scores where only one or a few verbs have been calculated. Then we identify the secondary gender of characters and compute the gender power diference of the alpha character and omega character.
Before analyzing the results, we close read a sample of 40 stories to assess the efectiveness of the verb lexicon-based method. This selection balanced time and gender power score variations, including five stories from extremely positive and negative representatives and earlier period’s representatives. Two human evaluators independently read and assessed each story. They then discussed their assessments to reach a consensus. Given the fragmented nature of fanfiction narratives and the broad concept of gender power diferences, we used a simplified three-category classification: positive diference, no significant diference, and negative diference.
As shown in Table 2, after excluding one fanfiction that used Omegaverse-related tags but did not adhere to the trope’s conventions, 27 out of 39 stories had human judgments that aligned with Riveter’s results, yielding an accuracy of 70%. In three stories where Riveter indicated a negative gender hierarchy, close reading revealed that despite more active descriptions of the Omega character, the actual narrative still portrayed the Alpha as dominant and the Omega as passive.
Additionally, we evaluated Riveter’s verb-based recognition accuracy, finding it could accurately capture most actions of the characters. However, due to limitations in coreference resolution, it could not detect finer details of bodily movements, such as ”his arm” and ”his hand.” Despite these limitations, the manual validation demonstrated a high consistency between the connotation frames-based gender power diference measurement and human judgment. The data quality of the Omegaverse corpus was also afÏrmed as reliable.
Consequently, based on these computational results, we analyzed the evolution of gender power diferences across various fandoms within the Omegaverse. The final result based on 3,468 fanfiction works in the Omegaverse corpus is shown in Table 3, and the change over time is presented in Figure 2.
First, gender power diferences of Alpha and Omega characters in slash relationships are broadly exhibited across diferent fandoms. Despite many stories in each fandom receiving negative gender power diference scores and close readings confirming the subversion of traditional hierarchies and promotion of gender equality, for all fandoms the gender power diferAlpha/Omega Gender Gender Riveter Power Dif- Power DifScore ference ference Std Amount Average
Unique Author Rate (%) ence scores approximate normal distribution with a mean above zero. This means that most of the times authors use the Omegaverse trope consistently with its definition. Moreover, Western TV and movie-related fandoms generally have higher scores than Japan Anime and K-poprelated fandoms. The fanfiction stories about the Will Graham/Hannibal Lecter pairing from Hannibal show the highest average and variance in gender power diferences, likely due to the series’ inherent violent and extreme elements. In contrast, pairings like Katsuki Yuuri/Victor Nikiforov fromYuri on Ice and Jeon Jungkook/Park Jimin from BTS, both of which have the smallest mean and variance in gender power diferences, highlight how specific fandom characteristics influence these dynamics. Yuri on Ice being a female-oriented romantic anime, shows stable and low gender power variance, echoing the original plot. Similarly, BTS pairings, despite high activity in recent years, maintain low mean and variance scores, reflecting a consistent portrayal within the K-pop fanfiction community.
This leads to the second common phenomenon we noted: most fandoms exhibit more withingroup consensus when more fans start writing. Initially, when only a few people created Omegaverse-themed stories, there was a higher average gender power diference. As more fans began writing, the average gender power diference significantly decreased, and the variance correspondingly narrowed. The fanfiction published in 2010 with the Castiel/Dean Winchester pairing fromSupernatural is the earliest Omegaverse story in this corpus, showing a significantly higher-than-average gender power diference score, which close reading supports. Similar patterns occur inMy Hero Academia. However, as more stories are created, the two pairings fromMarvel Cinematic Universe demonstrate stable low gender power diferences and low variance during the period between 2019 and 2020 when there was a active fanfiction creation (for example, for the Steve/Tony pairing, there were 37 stories in 2019 and 63 in 2020, compared to just 22 stories over the four years from 2012 to 2015). Similarly, Bakugou/Midoriya and two K-pop-related pairings share a similar pattern from 2019 to 2022, indicating that when a new trope becomes widely popular, authors and readers form a consensus over time, leading to consistent portrayal patterns within these fan communities. Initially, fans creating the Omegaverse trope may prefer dramatic gender hierarchy settings, but as the trope gains popularity, more authors use it as an opportunity for reflection. Additionally, the periodic fluctuations observed in Western TV series and film universes suggest that the mean and variance of gender power diferences may vary cyclically, potentially influenced by changes in fan engagement and preference developments within fan communities.
Overall, these findings underscore the widespread presence and relative stability of gender power diferences in Omegaverse fanfiction. This study highlights the dynamic interplay between fan engagement and gender power portrayal in these narratives, ofering insights into how fan communities evolve and challenge traditional power structures.
Our study uses an Omegaverse corpus that includes only English short to medium-length fanfiction from the top fandoms, in which multiple fanfiction works may written by the same authors. Additionally, by comparing gender power diference computation results across diferent fandoms, we observed a possible interaction between the original plot and power dynamics. This suggests that the original plot and narrative genre can bias the power dynamics, as seen in the Hannibal and Yuri on Ice examples.
In the next step, we plan to create a corpus with more variety in terms of fanfiction language, fandom plots, and themes, while controlling for author uniqueness. Additionally, we will also incorporate more nuanced linguistic features, such as adjectives and adverbs, to enhance our analysis. Tags potentially related to gender dynamics, such as ”Non-Traditional Alpha/Beta/Omega Dynamics” will also be systematically integrated into the evaluation process.
This work is part of the Graphs and Ontologies for Literary Evolution Models (GOLEM) project, a 5-year (2023-2027) research project funded by the European Commission (ERC StG). We would like to thank Junyao Zhang for participating in the evaluation procedure.