A Powerful Hades Is an Unpopular Dude. Dynamics of Power and Agency in Hades/Persephone Fanfiction

4.1 Data

A dataset of fanfiction was collected using the AO3-Scraper (Radiolorian 2022). The metadata is described in MythFic Metadata (Neugarten and Smeets 2023a, b). The 5,154 stories in the dataset all originate from Archive of Our Own (AO3), and were published between the platform’s inception in 2008 and the data collection in 2022.¹ All stories were tagged by AO3-users as belonging to the fandom Ancient Greek Religion and Lore, although the set also contains some overlap with other fandoms, both those related to Greek mythology (such as the popular young adult book series Percy Jackson), and those unrelated to Greek mythology (such as Sherlock Holmes and Harry Potter fandom). All stories were written in English. From the MythFic dataset, I selected stories tagged by their authors with the Hades/Persephone relationship and meeting the length criteria described in section 3, resulting in a dataset of 482 short stories. Descriptive statistics for the dataset are given in Table 1.²

4.2 Method

This paper analyzes connotation frames, a concept first introduced by Richard Fillmore (1976), to describe the conceptual frames that words evoke. For instance, the verb ‘exchange’ implies the existence of a giver, a taker, and a good being exchanged. Connotation frame analysis relies on the assumption that by connecting entities together through a predicate, texts “subtly connote a range of implied sentiments and presupposed facts about the entities” (Rashkin et al. 2016, 311). For example, the sentence “He violates her” casts the entity “he” as a perpetrator and “her” as a victim, and may evoke sympathy or pity. It may even imply that the female entity is valuable or desirable. Connotation frame analysis uses the meanings or connotations implied in agent-verb-theme relationships to assess such linguistic framing of dynamics between textual entities.

In this paper, I use Riveter (Antoniak et al. 2023) to measure the power and agency of two entities, ‘Hades’ and ‘Persephone’, on a large scale. Riveter parses texts, detects and clusters entities, extracts agent-verb-theme triples and matches these against a pre-selected lexicon of connotation frames – in this case, lexicons for power and agency (Sap et al. 2017) – to assign the extracted entities scores in the relevant semantic dimensions. It is worth noting that Riveter’s scoring system can lead entities to be assigned positive or negative scores, because different connotation frames can add or subtract scores. For example, the verb ‘defeat’ increases an entity’s power in relation to a theme, while ‘apologize’ decreases it. Because scores are aggregated at the level of entire stories, it is also possible for these scores to cancel each other out so an entity ends up with a power score of zero. While power is always calculated through these connotation frames as relational – if I defeat you, my power is increased in relation to yours – agency is not. The agency of an entity can be increased or decreased depending on the verbs they are associated with, independently of other entities in a text. For example, ‘managing’ increases an entity’s agency, while ‘waiting’ decreases it.

Advantages of Riveter include its ease of use and the interpretability of its results. Compared to other lexicon-based tools, Riveter also has the benefit that its pipeline takes grammatical structure into account. One important drawback of Riveter, or any lexicon-based tool, is that the results are only ever as good as the lexicons being applied, and so domain-specific indicators of power or agency may be overlooked. This is perhaps especially troubling since fanfiction presents an online and in some ways subcultural domain of language use. In previous research applying Riveter to fanfiction-texts (Neugarten 2025), a small-scale manual error evaluation conducted by a single annotator found the tool’s scoring accurate 57% of the time in detecting power and 89% of the time in detecting agency. Most errors had to do with an inability to detect metaphorical language use and failing to account for the ambiguity of power dynamics in some instances. On the other hand, the connotation frames for power and agency applied here have been shown to perform well on contemporary movie scripts (Sap et al. 2017), a domain that is somewhat similar to language use in contemporary online fanfiction.

Another drawback of the Riveter tool is that it reduces gender to a binary variable (male versus female) and assumes that each detected Hades-entity is male while each detected Persephone-entity is female. Fanfiction communities have a relatively large contingent of participants whose gender identity goes beyond the binary (Rouse and Stanfill 2023), and fanfiction correspondingly explores and represents the experiences of fictional characters with nonbinary gender identities more often than mainstream fiction does, even if these explorations are not always explicit (Leetal 2022). However, in MythFic Metadata, tags describing gender identities beyond the binary were highly infrequent and not explicitly linked to Hades or Persephone. Unfortunately, distant readings (almost) always trade in a measure of granularity or specificity to gain a broader view, and I find the tradeoff acceptable in this case because of the low frequency of tags indicating the presence of nonbinary genders in the stories.

5.1 Power and Agency Scores

Table 2 provides descriptive statistics of power and agency scores per entity, aggregated over the entire corpus. Following previous research that applied Riveter to fanfiction texts (Yang and Pianzola 2024), I also used Riveter scores to calculate the power differences between the two relevant entities – Hades and Persephone – for each individual story. I also added a comparison of agency scores between the two characters. For both the power and agency dimensions, these difference scores – for which descriptive statistics are also listed in Table 2 – were calculated as follows:

This means that a positive difference indicates that Hades scored higher on power or agency, while a negative difference indicates that Persephone scored higher.

The average scores indicate that both characters have negative power scores, with Persephone (-0.09) scoring lower than Hades (-0.07). A t-test revealed that this difference was not statistically significant (t = 0.48, p = 0.63). In 37 instances, Persephone was assigned more power, while in 29 instances Hades was assigned more power, and in three instances they were assigned equal amounts of power, canceling each other out and leading to a power difference of zero. Both characters have positive agency scores, with Hades (0.22) scoring slightly higher than Persephone (0.21). Again, the difference was not statistically significant (t = -0.54, p = 0.59). In 35 instances, Persephone was assigned more agency, while in 33 instances Hades was assigned more agency, and again the scores were equal in three stories, canceling each other out. When looking at the size of the differences, the negative mean power difference (-0.03) indicates that Persephone tends to outdo Hades when it comes to the power dynamics between them, while the positive mean agency difference (0.02) shows that Hades tends to outdo Persephone in terms of agency.³ Keep in mind that the power and agency difference scores are based only on those stories where both characters were assigned a power or agency score so that a difference between the two could be calculated. These difference scores thus represent a smaller subset of the corpus (only 69 and 71 stories respectively) than the entity scores.

5.2 Shifts over Time

Existing research on the way literature represents and may impact the social world, and gender dynamics in particular, has hypothesized that the position of female characters in narrative may have shifted over time with regards to empowerment and oppression because narrative may either reflect or shape the emancipation of women in the real world. However, this emancipatory hypothesis has been rejected, at least for Dutch-language literary novels (Smeets 2024), but also when it comes to increasing the real-world prominence of female authors (Underwood et al. 2018) and their prestige in literary circles (Koolen 2018). Existing scholarship thus seems to indicate that literature is not as progressive in terms of gender politics as academics may like to think.

Fanfiction, however, has a subcultural position outside the literary establishment, so it is possible that these stories do not adhere to the same patterns when it comes to representing gendered social dynamics as published literature. This raises the question: Does the fanfiction case study show a change in gendered dynamics of power and agency over time? On the one hand, it is important to note that the period covered in this fanfiction corpus (2007–2022) encompasses a relatively short span of time. On the other hand, during this time a lot has changed when it comes to the reception of the Hades/Persephone relationship in popular culture. At least 15 retellings of the Hades/Persephone myth aimed at young adults have been published during this period, including the immensely popular web comic Lore Olympus (Smythe 2021) which does not portray Hades as a domineering, oppressive, or abusive partner, but rather as a soft-spoken, attractive love interest. It is possible that the increasing popularity of these retellings has impacted the dynamics of power and agency portrayed between Hades and Persephone in fanfiction over time.

5.2.1 Power Shifts over Time

Figure 1 presents a plot of all power differences over time. Each blue dot represents the power difference assigned to a single text, with negative scores indicating that Persephone had more power and positive scores indicating that Hades had more power. The red line indicates the 1-year rolling mean of the power difference. Around 2012, several stories were published with a marked power difference in favor of Persephone. This is interesting, although it is perhaps too few to speak of a true trend. Around 2019, stories started exhibiting a power difference favoring Hades. In this period, fanfiction production about Hades and Persephone increased markedly overall (see Figure 2), perhaps because of the release of the popular video game Hades (2018). In this game, Hades is initially the main antagonist, which may partly explain the increasing fanfiction production with a power difference favoring him. This trend indicates that fanfiction is not empowering its central female character, Persephone, more and more over time. In other words, the development of power differences is not progressive when it comes to representing gender equality, if we take progressive to mean disadvantaging the female character less and less over time. Nonetheless, the power imbalances hovering around zero that we see in Figure 1 may be considered transformative in light of the strikingly unequal and disempowering dynamic that is characteristic of more culturally dominant versions of the Hades/Persephone myth.

Figure 1: Plot of power difference of Hades and Persephone over time. Negative scores indicate that Persephone had more power; positive scores indicate that Hades had more power.

Figure 2: Histogram of publication dates.

5.2.2 Agency Shifts over Time

Figure 3 presents a plot of all agency differences over time. Each blue dot represents the agency score difference between Hades and Persephone calculated for a single text, with negative scores indicating that Persephone was ascribed a higher agency score by Riveter, and positive scores indicating that Hades was ascribed more agency. The red line indicates the one-year rolling mean agency difference between the two entities. From 2013 to 2015, there was a pattern of decreasing agency difference scores over time, which is to say that Persephone’s average agency relative to Hades increased over that period. However, around 2016 and again around 2018, stories with an agency difference favoring Hades disrupted this trend. In recent years, these mean difference scores have hovered around the zero-line, indicating no strong inequality when it comes to the distribution of agency between Hades and Persephone.

Figure 3: Plot of agency difference of Hades and Persephone over time. Negative scores indicate that Persephone was ascribed a higher agency score; positive scores indicate that Hades was ascribed more agency.

5.3 Comparison to Existing Metadata: Does Power or Agency Correlate with Story Violence?

It is perhaps to be expected that those stories where Riveter detects a large difference in the power or agency between characters are also more likely to contain incidents of violence. After all, violence can be understood as an extreme – usually physical – exertion of power by one entity over another. Violence can also have the effect of limiting the agency of its victim. Fanfiction writers on Archive of Our Own tend to attach detailed and accurate content-oriented tags to their stories, making it easy for readers to find stories that fit their tastes and to curate their reading experiences with regards to potentially undesired themes or topics. This tendency to attach detailed metadata (tags) to stories gives scholars in fan studies valuable insight into the story-level content of fanfiction, although these tags are not always very fine-grained. In MythFic Metadata, for example, the 5,154 stories for which metadata is provided are accompanied by 1,3936 unique additional tags. A little more than 600 stories in the dataset are not tagged with any additional tags, but for those stories that have been tagged, the tags often provide valuable content-level information about genre classifications, plot elements, characterization, or story world characteristics. Previous research has used these tags to measure correlations between different types of romantic relationships (including straight, lesbian, and homosexual) and the presence of violence in fanfiction (Neugarten 2024). Using the same tag-based operationalization of violence, I was able to calculate correlations between Riveter’s power and agency scores and the presence of tags indicating violence. Because the presence of violence in stories was indicated as a binary variable (1 for violence, 0 for no violence), I calculated Spearman’s ρ correlations to be able to compare this binary data with the Riveter scores.

I used four of the five types of violence identified in previous work (Neugarten 2024): physical violence, sexual violence, captivity, and death. The specific tags used as a proxy for the presence of these different types of story violence are listed under supplementary materials (Appendix A). I disregard the fifth category of violence – roughness – because evidence suggests that the tags related to this category (rough sex, biting, hair-pulling, and spanking) were not reliable indicators of violence that entail unequal dynamics of power and agency. Instead, these tags were often used to describe consensual sexual acts.

Figure 4 presents Spearman’s ρ correlations (with a significance threshold of 0.05) between the presence of violence as indicated by tags and Riveter scores.⁴ It is unsurprising that the two characters’ power and agency scores correlate with the power and agency differences between them,⁵ and that some types of violence correlate positively with each other, such as physical violence with death (0.17) and sexual violence with captivity (0.28). This final correlation is nonetheless interesting, because it points to the existence of stories that reflect the dominant canonical version of the Hades/Persephone myth, in which Persephone is abducted and raped, in the dataset.

Figure 4: Correlations (Spearman’s ρ) between violence and Riveter scores

Fanfiction featuring death has a weak positive correlation with Persephone’s power scores (0.11) which raises the question of who is dying in these stories. Surprisingly, Persephone’s agency scores are slightly positively correlated with the presence of non-consensual sex acts (0.10). This is counterintuitive for two reasons. Firstly, I expect that Persephone is more often the victim than the perpetrator of sexual violence in the fanfiction corpus because that is also the dynamic most prevalent in the myth’s culturally dominant version. Secondly, I do not associate the narrative or semantic role of being the victim of sexual violence with a high level of agency.

Closer inspection of the data shows that 13 stories in the corpus were tagged with sexual violence. Of those, four were not assigned any Riveter scores. The Riveter scores for the remaining nine stories are listed in Table 3. It then becomes evident that Persephone is assigned positive agency scores in six of the stories and positive power scores in three. Some of these stories call into question the assumptions underlying the metadata analysis. Firstly, in some instances, tags indicating sexual violence that is ‘implied’ or ‘past’ – but not literally described in the narrative text – account for the correlations. In other instances, the direction of sexual violence is not aligned with what one would expect to see in a patriarchal story world, for example, when Persephone is perpetrating sexual violence against Hades instead of vice versa. These stories, in which the expectations of gendered inequality are reversed or the past occurrence of sexual violence is used as a narrative basis for a story of empowerment, may be interesting candidates for close reading in future research.

5.4 Comparison to Existing Omegaverse Analysis: Does Gendered Power Behave Differently than A/B/O Power?

Previous research (Yang and Pianzola 2024) examined the gendered power dynamics of the Omegaverse, a popular subgenre of fanfiction. In the Omegaverse, the culturally dominant division of gender as a male/female binary is expanded because characters have a secondary gender. For this secondary gender, three options exist: Alpha (the dominant gender), Omega (the submissive gender), and Beta (the neutral gender). For this reason, the Omegaverse is also often referred to as A/B/O, an abbreviation of these three genders. In most Omegaverse stories, A/B/O hierarchy dictates how people interact in erotic and romantic situations, but in some versions their entire social world is structured around this hierarchy. As noted by Milena Popova, “an alternate universe where gender and sexual scripts work radically differently to ours, such as the Omegaverse, is the perfect tool to explore the effect of scripts and dominant ways of thinking on our actions and our ability to meaningfully negotiate consent” (2021, 58). In other words, the imagined society of the Omegaverse can be a way for fanfiction to engage transformatively with dominant cultural ideas structuring gendered power relations. This also makes it interesting to compare gendered power dynamics between fanfiction that takes place in the Omegaverse and fanfiction in story worlds that are more similar to real-world societies in their conceptions of gender. At the time of writing, Archive of Our Own hosts over 261,000 stories tagged with this trope, which shows its popularity.

Because Omegaverse stories imagine gender and its associated power dynamics in a way that is radically different from real-world societies, Yang and Pianzola (2024) decided to map the gendered power dynamics between selected Alpha and Omega characters in a dataset of Omegaverse fanfiction. They found that gender power difference between Alphas and Omegas can be more-or-less stably detected over time within particular fandoms, although “most fandoms exhibit more within-group consensus when more fans start writing” (Yang and Pianzola 2024, 914). Although the Hades/Persephone dataset used here is smaller than the multi-fandom dataset used by Yang and Pianzola (2024), Figure 5 and Figure 6 suggest a similar trend in this use case, with both power and agency differences fluctuating less as more and more fanfiction is written in the fandom year by year.

Figure 5: Average power difference and number of publications over time. Negative scores indicate that Persephone had more power; positive scores indicate that Hades had more power.

Figure 6: Average agency difference and number of publications over time. Negative scores indicate that Persephone was ascribed a higher agency score; positive scores indicate that Hades was ascribed more agency.

Comparing gender power difference between the Yang and Pianzola (2024) study and the case study presented in this paper presents one methodological difficulty: Because Persephone and Hades always have the same gender in the corpus used here, I have simplified the calculation of gender power difference to account for this, so that negative power differences indicate a positive power difference for Persephone. Conversely, in the dataset used by Yang and Pianzola (2024), characters can have different A/B/O genders in different stories. In their research, a positive gender power difference always indicates more power for the Alpha (the traditionally dominant partner), while a negative gender power difference means more power for the Omega (the traditionally submissive partner). In terms of the real-world construct of the gender binary, all characters included in their dataset are male.

Table 4 provides descriptive statistics of the gender power differences from the Omegaverse (Yang and Pianzola 2024), with the same statistics for the Hades/Persephone case study for comparison. Since Yang and Pianzola did not measure agency, no comparison can be made on that variable.

Compared to the A/B/O relationships analyzed by Yang and Pianzola, then, we can conclude that the average power difference between Hades and Persephone is small and slightly in favor of what may perhaps be called the traditionally less powerful partner: the young girl Persephone. In contrast, all average differences calculated by Yang and Pianzola (2024) result in positive scores, pointing to a power difference in favor of the traditionally dominant Alpha. It is worth noting that all relationships analyzed by Yang and Pianzola are slash – romantic or sexual relationships between men. In comparison to their dataset, then, it seems that the Hades/Persephone fanfiction analyzed in this paper is slightly less unequal, if we take equality to mean that women are portrayed as powerful main characters in relation to their (male) love interests. On the other hand, the Hades/Persephone relationship has a higher standard deviation than most relationships examined by Yang and Pianzola, suggesting that the gender power difference between them is more variable than for other relationships.

One possible explanation for these observed differences is that the Omegaverse fanfiction studied by Yang and Pianzola (2024) relies for many of its genre conventions on a rigid hierarchy of genders, with Alphas (almost) always portrayed as dominating over Omegas in every aspect of their interpersonal, sexual, and social interactions. In these genre conventions, the Omegaverse differs from the corpus of Hades/Persephone stories. If interpersonal dynamics between Hades and Persephone are more flexible in relation to genre conventions than between Alphas and Omegas, this explains the higher standard deviation in their power difference scores.

5.5 Does Power or Agency Correlate with Common Popularity Metrics on AO3?

As described in subsection 4.1, Archive of Our Own also provides statistics on various popularity metrics for fanfiction and these metrics are included in MythFic Metadata. The different kinds of popularity-related metrics provided on AO3 are listed and described in Table 5.

Each of these popularity metrics indicates a different kind and level of engagement. A hit does not necessarily mean that a story has actually been read. Kudos are quickest and easiest to give, and while they communicate a positive evaluation or encouragement to the author, this may not be as engaged as bookmarking a story to revisit it later, which has been described as a “stronger and stickier form of approval than a simple ‘kudos’” (Vadde and So 2024, 24). Finally, a typed-up comment may be considered the strongest indicator of approval. In a previous study on fanfiction comments, I observed that “commenters can be characterized as above-averagely engaged or committed readers, since they invest the time and effort to comment” (Neugarten et al. 2024, 2020). A 2013 census of Archive of Our Own also found that only 43.6% of the platform’s users regularly leave comments (centreoftheselights 2013). In this sense, commenters are not necessarily representative of fanfiction’s readership as a whole. Finally, hits are perhaps best understood as indicators of what readers find appealing at first glance or based on a brief description, while kudos, comments, and bookmarks are indicative of what readers appreciate after (mostly) reading the full text of a story.

Figure 7 presents Spearman’s ρ correlations (with a significance threshold of 0.05) between popularity metrics and Riveter scores. It is unsurprising that popularity metrics correlate strongly with each other. However, it is notable that Hades’ power scores have a mild negative correlation with three of the popularity metrics: -0.14 for kudos, -0.11 for bookmarks, and -0.12 for hits. In other words, stories that represent Hades as more powerful are slightly less popular. The power difference between Hades and Persephone also has a mild negative correlation (-0.10) with both kudos and hits, indicating that greater inequality in the power relation between the characters is less popular among readers. Persephone’s agency correlates positively with many measures of popularity: 0.16 for kudos, 0.15 for bookmarks, and 0.11 for hits, so stories that portray her as having more agency tend to be more popular. Hades’ agency also correlates positively with kudos (0.11), though not with other popularity metrics.

Figure 7: Correlations (Spearman’s ρ) between popularity metrics and Riveter scores.

When interpreting these correlations between Riveter scores and story popularity, time is a confounding factor. It is difficult to address this issue for two reasons. Concerning the Riveter scores, the writing and reading of fanfiction is subject to community trends that are complicated to track but that may well influence the production and reception of stories’ portrayal of power dynamics. In other words: The things that Riveter is measuring are context-dependent and likely to change over time. For example, recall the publication years of the stories in the corpus (Figure 2) and the notable increase in fanfiction production around 2019 that I hypothesized is tied to the release of the video game Hades (2018) (subsubsection 5.2.1). If Hades leads new audience members to enter the fan community with conceptions of characters and character dynamics shaped by the video game, this development may also influence how they portray these character dynamics in fanfiction, and which types of dynamics they appreciate. Closer inspection of the correlation data by year shows that a small subset of stories in 2012 and again in 2020 seems responsible for the negative correlations found between popularity metrics and Hades’ power scores.⁶ These stories are good candidates for future research, particularly close reading.

Regarding the popularity metrics, it is important to note that these metrics are also not independent of time. On the one hand, stories are likely to receive many of their hits, kudos, and comments soon after their date of publication because AO3 does not employ an algorithm. Instead, the most recent stories are always shown first in users’ search results. On the other hand, older stories have had more time to accrue kudos, hits, and comments. To address this difficulty, Jacobsen and Kristensen-Mclachlan (2025) use the metric of the kudos/hits ratio. This metric represents the proportion of readers who appreciated a story (kudos) relative to the number of readers who clicked on it in the first place (hits).

Correlations between the kudos/hits ratio and the Riveter scores for the Hades/Persephone corpus were mostly non-existent, both at the level of the entire corpus and when separating out stories by year of publication. However, taking all stories together, there was a slight negative correlation (-0.12) between the ratio and Hades’ agency scores, suggesting that a portrayal of Hades who is ascribed more agency is less popular among readers.

Despite the caveats related to the impact of time on both Riveter scores and popularity metrics, the correlations presented in Figure 7 do generate some insight into how power and agency circulate in these stories. For both characters, their own levels of agency and power do not correlate with each other, although Persephone’s agency has a weak positive correlation (0.16) with Hades’ agency, suggesting that some stories ascribe more agency to both characters, and the agency of one need not come at the expense of the agency of the other. Interestingly, there is no correlation between the power scores for the two characters, suggesting that power dynamics are not actually a zero-sum game. Persephone’s power does have a weak negative correlation with Hades’ agency (0.09), suggesting that his ability to take action in a particular story world sometimes limits her power in that universe. Furthermore, it stands out that the power difference between the two characters correlates positively with the agency difference (0.26) – when the dynamics in a story are more unequal in terms of power, this correlates with more unequal dynamics in terms of agency.

6.1 Discussion

Two main areas present themselves as fruitful for future research. Firstly, it is of course possible that gendered power dynamics function differently in longer stories, especially because their buildup of narrative tension may operate in a different way than a short or very short story does. Future research may therefore want to turn attention to those longer stories about Hades and Persephone – or other characters from mythological narratives or popular fiction – that have not been covered in this paper.

Secondly, as noted in previous research (Neugarten 2025), it is difficult to draw fine-grained conclusions based on Riveter scores without contextualizing and evaluating these scores through close reading. Although detailed close readings fall outside the scope of the current paper, this also presents an interesting avenue for future research.

6.2 Conclusion

To conclude, let me address the research questions raised in section 2 one by one.

• In short-form fanfiction about the relationship between the mythological characters of Hades and Persephone, how is their power and agency portrayed?

  Both Hades and Persephone are portrayed in fanfiction about their relationship as relatively disempowered because Riveter assigns both entities negative power scores on average (-0.07 for Hades and -0.09 for Persephone). The average difference between the two characters is not large (-0.03). For agency, both characters are assigned low but positive scores (0.22 for Hades and 0.21 for Persephone), and the average difference is even smaller (0.02).

• Is their power dynamic gendered, i.e., does the gender of these characters impact their level of power or agency in the stories?

  In stories where both characters are assigned scores in a given category (meaning that a difference could be calculated), Persephone tends to score higher than Hades on power, while Hades tends to score higher than Persephone on agency. This indicates that something transformative is going on in this corpus of fanfiction: The character with many identity characteristics traditionally associated with weakness or disenfranchisement – youth and femininity – tends to have the power difference in her favor in this small subset of the corpus (n = 69).

• How does this dynamic shift over time?

  Around 2019, stories became marked by a mean power difference favoring Hades. This roughly coincides with increased fanfiction production focusing on the Hades/Persephone relationship. Between 2013 and 2015, there was a pattern of decreasing mean agency differences over time. In other words, Persephone’s average agency relative to Hades increased over that period. After that time, this pattern was no longer discernible. It is an interesting focus for future research to look at how these patterns have developed since 2022, the year of data collection for MythFic Metadata. Since then, the total number of works of fanfiction in the Ancient Greek Religion and Lore fandom on Archive of Our Own has almost tripled, from 5,154 to 14,850 at the time of writing, and the total number of stories about the Hades/Persephone relationship in that fandom has increased from 844 to 1,185.

• How do power and agency scores compare to existing research (Neugarten 2024; Neugarten and Smeets 2023a, b) on violence and gendered power dynamics in fanfiction about Greco-Roman Antiquity?

  It was not surprising to find positive correlations between the different types of violence under analysis. It was surprising, however, that stories tagged with ‘death’ showed a weak but positive correlation (0.11) with Persephone’s power scores, suggesting either that the death in many cases may not have been her own or that death in the context of becoming Hades’ partner and the queen of the Underworld can be an empowering experience. It was also surprising to find that stories tagged with non-consensual sex showed a weak positive correlation (0.10) with Persephone’s agency scores, prompting a closer examination of what was going on in those stories.

• How do these scores compare to existing research (Yang and Pianzola 2024) on power dynamics in Omegaverse stories, a popular subgenre of fanfiction that presents a speculative conception of gender with clear power hierarchies?

  Compared to the power dynamics present in the Omegaverse fanfiction studied by Yang and Pianzola (2024), the power difference between Hades and Persephone was relatively small. The average power difference between Hades and Persephone was -0.03, while the scores reported by Yang and Pianzola – which reflect the gender power difference between the fictional genders of Alpha and Omega rather than those between men and women – ranged from 0.05 to 0.29. It thus seems that the fanfiction analyzed here was less unequal than the stories about the various Alpha/Omega relationships studied by Yang and Pianzola, provided we take Riveter scores as indicative of gender power difference and consider a smaller gender power difference to be less unequal than a larger difference.

• Do differing depictions of gendered power relations in this corpus of fanfiction impact the popularity of the stories among their readers?

  What stands out in this regard is that a higher power score for Hades is negatively correlated with a number of popularity metrics if we look at the corpus as a whole and disregard the role of time. These stories receive fewer hits (-0.12), suggesting that fanfiction readers are less likely to click on them. They also receive fewer kudos than other stories (-0.14), suggesting that readers are less likely to compliment these stories’ authors and express their enjoyment, and are bookmarked less often, suggesting readers are less likely to want to revisit these stories. Conversely, stories’ popularity correlates positively with Persephone’s agency score. This applies to hits (0.11), suggesting these stories are more often clicked on; kudos (0.16), suggesting these stories are more often liked; and bookmarks (0.15), suggesting these stories are more often revisited.

What, then, can we conclude about fanfiction’s capacity to transform the culturally dominant, gendered power dynamics of the Hades/Persephone myth, and, perhaps more broadly, its capacity to transform culturally dominant discourses related to gender, but also other topics, through rewriting?

On average, we see power differences favoring Persephone and agency differences favoring Hades. This suggests that there is a pattern of difference structuring the distribution of power and agency between these two characters, although more research would be needed to determine whether these patterns of difference can be considered representative of each character’s gender. In comparison to the (fictional) genders Alpha and Omega, power differences between these two (male and female) characters are small. Trends in the dynamics of power and agency between Hades and Persephone became visible over time and point to some interesting avenues for future research.

The most interesting finding, perhaps, is that stories with more power assigned to Hades are significantly less popular, while stories with more agency assigned to Persephone are significantly more popular. However, this trend, which emerges when examining the corpus as a whole, cannot be detected when examining data year by year. In the end, fanfiction is a reflection of the kinds of stories fans are most interested and invested in at a particular time. In these correlations, we can see a desire from the fanfiction-reading audience for gendered power relations to be less unequal. It is clear that fanfiction has the capacity to be transformative of unequal power dynamics between male and female characters like Hades and Persephone, though not all stories are. The stories that are transformative, however, are often rewarded with more appreciation and engagement from their readership.