1. Introduction
[1.1] Since late 2022, artificial intelligence (AI) image and text generation tools (referred to throughout as generative AI) have risen in popularity. These softwares use machine learning models to produce art and writing seemingly without human intervention (production is entirely automated, though humans are required to provide training data, train models, and provide prompts). Widespread public use of AI-generated imagery and text has prompted backlash (Beckett and Paul 2023; Roose 2022), labor strikes (Collier 2023), and legal challenges (Vincent 2023; Woods and Ma 2023; Bruell 2023). Proponents argue that these tools allow for greater democratization of art and are justified under fair use interpretations of copyright law (Vincent 2023; Wong 2022), while opponents cite data privacy, ethical, and intellectual property concerns (Dixit 2023; Edwards 2022).
[1.2] For fandom, AI presents a unique set of challenges. Machine learning models that power programs like ChatGPT have allegedly taken large amounts of text data from fan fiction websites like Archive of Our Own (Eveleth 2023), and many fan conventions have announced bans on selling AI-generated works after backlash from attendees (Mattei 2022). Though empirical studies about fans' views on AI are still in their infancy, it is clear this issue is, at minimum, divisive; one can look at the comments of a post from 2023 by the Organization for Transformative Works discussing AI-generated works and the Archive to see many strongly opinionated fans with many different positions (Eskici and Organization for Transformative Works 2023).
[1.3] As fans and researchers, we believe the place of AI-generated works in fandom necessitates reevaluating existing core tenets of the field. Our purpose here is to meditate upon that place and to pose questions for academic studies of fandom to grapple with as this technology evolves. While we do not view AI as inherently problematic (in fact, we acknowledge the great strides made possible by machine learning technologies, particularly for academic research), we believe that current ethical dilemmas and systems of power create potential for widespread harm. We present a set of core questions that we believe academic fan studies will have to contend with as we move forward. We speculate on preliminary answers on the basis of prior understandings of fan cultures and fan-industry relationships and present one theoretical framework that we believe may guide future work.
[1.4] We see opportunities for academic dialogues around the following questions, many of which this special section will likely intervene on: (1) How does the use of generative AI in a fannish context dialogue with previous ideas of fan labor and production, by scholars and fans alike? (2) How might we conceptualize fandom praxis as a labor of love when, arguably, neither labor, nor love, is involved? (3) How does generative AI change existing typologies of fannish intersection with (and exploitation by) industry?
[1.5] The place of AI media in fandom spaces will no doubt continue to be hotly contested. Martine Mussies posited (2023, ¶ 5.1) that since 2023, "who made a particular image and how do not matter; the work functions in a [producer] community, so the image is not the end but a means toward it." Ultimately we disagree with this position and argue that how an image is made does matter, as these technologies exist within a structural system of exploitation that ultimately hurts fans and fan works. We take the position that generating work via AI software (where the only human input is typing a prompt into an algorithmic generator) is fundamentally distinct from human creative processes. This is recognized by governing bodies like the US Copyright Office (Brittain 2023), and we share their opinion that AI-generated works do not constitute human authorship, regardless of philosophical arguments about computers' creative ability. Fan studies emphasizes the personal, valuable nature of creative practice, and we believe automating this practice via tools built upon capitalistic logics of datafication and exploitation of creative works as fodder for machines is incompatible with logics of fandom cultures.
2. Fan creation as a labor of love
[2.1] Fan studies research emphasizes fandom's noncommercial gift economy. Fannish work by nature is unpaid and undervalued (Anselmo 2018), operating on a model of reciprocal exchange. Fannish economies of production center what Tisha Turk (2014, ¶ 2.1) describes as a "labor theory of value...gifts made by fans for fans. The worth of these gifts lies not simply in the content of the gift, nor in the social gesture of giving, but in the labor that went into their creation." Fandom also operates on a collaborative model for projects (Jones 2014) and emphasizes reciprocal exchange (Hellekson 2009).
[2.2] Indeed, Busse (2015, 112) argues that "fans tend to regard fan labor as a labor of love and as a shared passion —and, in many cases, as paying it forward" in an era where media and entertainment industries often exploit fans' work. The labor of love type of gift work is therefore predicated on creation, where fans do not expect to receive financial compensation in return. Mark Duffett recognizes that fandom's affective dimension, where emotional engagement drives fans to seek out popular culture texts, "is often used as a form of shorthand to express individual humanity" (2013, 90). In this way, the gift work reflects a fan's interpretation of the original text, their understanding of fandom discourse, and their personal style —thereby reaching out to community members who hopefully feel and/or create similarly. Fandom's gift economy and its associated performances reflect a culture of reciprocity and respect where fans are expected to ascribe to the labor of love philosophy. While there are likely cases where fans wish to use AI as part of this exchange, such as when a fan fic writer might supplement their fic with AI images, for example, this scenario still centers human creation; AI does not replace the human fan work but augments it. AI becomes a tool within a human-led creative process rather than a tool to replace those processes. We would also note that in this situation, a fan might participate in an exchange or commission an artist to keep community norms central and thus maintain the exchange culture of fandom.
[2.3] One might ask why AI's (alleged) theft of data and artistic works is distinct from fans' so-called theft of texts, famously conceptualized by Henry Jenkins as poaching of texts in his 1992 titular work Textual Poachers. We note a difference in both motivation and outcome; while fans' theft is done to uplift, celebrate, and build upon original texts (with demonstrable generation of value for the original media; Lantagne 2015; Morgan 2021), AI's theft seeks to build a product that will replace those texts. A fan's work transforms the original text and does not compete with that text, while AI-generated media demonstrably has the effect of automating and replacing humans' labor (Carter 2023; CVL Economics 2024). The explicit profit motive is another distinction here; harvesting of fan works for capitalistic gain decontextualizes the fan experience and upsets the gift work economy. However, even in the case of open-source and/or not-for-profit AI platforms, the aforementioned differences in motivation and culture still persist.
[2.4] Fan works' affective nature is integral to the gift work economy because it suggests an emotional connection to the community. Tight-knit social media platforms, especially Tumblr, welcome and desire the reciprocal nature of fandom. Melanie Kohnen (2017, 339) claims of this space, "Fannish affect at large is a complex and self-reflexive engagement with media" that is misunderstood by industry because of its many rules of performance. We only have to look at how fans and scholars recognize fan work production as a labor of love in the first place as a signifier that emotional attachment —to text and fandom alike —are essential to fannish experience.
[2.5] It is difficult to see how generative AI provides the kind of value that Turk, Duffett, and others describe. AI also calls into question existing frameworks of fans' affective engagement with texts; how is this kind of reflexive engagement possible if creative processes become automated via capitalist, profit-driven technology platforms? If, as Lucy Bennett and Paul Booth argue, "in the digital space, everything fans post, create, or share could be considered a type of performance," then there are specific actions a fan can perform in order to show how dedicated they are to their fan community (2015, ¶ 1.4). This list can include anything, from more traditional fan works like fan fiction and fan art, to the less so such as cosplay and video edits, or fan vids. These performances reflect a desire to engage with fandom emotionally (putting one's self into one's works) and the gift aspect suggests a desire to give back to a supportive, valued community.
[2.6] In our view, these types of performances are not possible if labor and creation is functionally bypassed to achieve an end product. These technologies are built upon uncompensated repurposing (and, to many, theft [Ho 2023]) of creative works, and this structural system of exploitation is incompatible with creative ideologies and praxis of fan communities. Automating creative processes via text and image generation tools provides a fundamentally antithetical logic to the kinds of performances and creative practices that fan community norms value, even, we argue, in cases where generative AI is used as supplement rather than a replacement for fan works. It further perpetuates exploitation of fans, fan culture, and fan works by industry.
3. Fans and industry: Media, technology, and exploitation
[3.1] As fandom has moved into the spotlight of mainstream popular culture, fan studies researchers have begun studying fans' interaction with industry. Whereas in prior years, fans worried about overly litigious creators such as Anne Rice (infamous in fan circles; Jackson 2021), today fan-industry scholarship is concerned with different questions. Academic research on fandom emphasizes how fans influence and are influenced by producers (Galuszka 2015), fannish intersection with industry space, such as at convention venues (Hanna 2014), and others. AI-generated media creates new questions still.
[3.2] Fan-industry relationships remain well debated. Mel Stanfill (2019) compared fans and industry to domestication of livestock, arguing that media industries seek to control fans and exploit their usefulness rather than creating a harmonious relationship. Suzanne Scott (2019) emphasized that industry is friendly to certain fans and views others (predominantly women fans and fans of color) as less desirable audiences. On the flip side, fans have been able to successfully influence industry as sponsors, stakeholders, and investors (Galuszka 2015) and gain input into media creation (Navar-Gill 2018), especially by generating value and advertising (Lantagne 2015; Morgan 2021).
[3.3] All sides of the fan-industry debate emphasize this constant tension between exploitation and influence. Kohnen (2017) writes that fans are often shut down by industry when they go too far, but studios simultaneously rely on fans for marketing and influencing conversation. The simultaneous valuation and devaluation of fan work by industry parallels discourses present in the rise of generative AI in the creative arts; one might argue that this technology was created by both valuing and devaluing arts simultaneously. Though artistic works are undoubtedly treated as valuable in that these programs are dependent on their inclusion in training datasets, it is also arguably true that the automation of art and writing creates subpar products, hurting human creativity and creating long-term devaluation of artistic works. Indeed, Ted Chiang (2023) famously described ChatGPT as creating "a blurry JPEG of the web," and other industry voices have noted that desires for AI to replace and/or scare artists stems from underestimating (and thus undervaluing) the creative process (Clarke 2022). This tension between valuable product and exploitable resource is not unlike the tension between fans and industry: You are valuable until you are not.
[3.4] We view AI-generated media as the latest in a long tradition of thorny problems for fans and fandom researchers to grapple with. We argue that this technology places online fans and fan works in a further exploitative relationship through the lack of consent inherent in the creation of (often proprietary) training datasets. While some argue that AI training is covered by fair use doctrine, it is unclear to what extent this will hold up in ongoing court cases given evidence of direct copying of works contained in training databases (Somepalli et al. 2022; Carlini et al. 2023). Given fandom's emphasis on creative practice and production of original, generally noncommercial works, we believe fannish attitudes towards these technologies will mimic those of other moments of perceived exploitation by industry actors.
[3.5] Historically, fans have not reacted well to these moments. FanLib, Kindle Worlds, and projects like it were received disastrously by fans (Hellekson 2009; Stanfill 2013) when industry actors attempted to capitalize on fannish gift economies. Many of these ventures do not respect fan community norms (Stanfill 2017), instead sticking to capitalist logics. In fact, these attempts at commercialization showcase what researchers describe as repackaging fandom and fan labor to generate capital (Scott 2009). While the processes involved in generative AI development are not the same as those of FanLib or Kindle Worlds (in that they are not explicitly asking fans to generate capital), they perform similar co-opting of fan works (via scraping for training datasets). Decontextualization is also relevant; like on Kindle Worlds and FanLib, fan works are removed from the fan audience and fed (literally, in this case) to a capitalist audience. Moreover (though empirical data is still forthcoming), we speculate that fans' reactions to fan works in training datasets will be similar to the reaction to the aforementioned platforms.
[3.6] As emphasized in our final section, we believe fan-industry scholarship must further incorporate the technology industry as well as the media industry. Considering structural, capitalistic ecosystems of AI development in relation to the arts (and by extension, to fan cultures/production) must become a priority. Prior work has already begun to examine digital fan cultures' evolution into an algorithmic, data-driven one (Yin 2020), and the place of fan cultures within a structural system of algorithmic capitalism will only grow more precarious as AI technologies continue to become more widespread. This is critical to interrogate as AI technologies in the arts evolve and reshape both the fan-industry relationship and the reciprocal, noncommercial aspects of fan spaces. In our final section, we present one theoretical framework that we hope will be useful for scholars of digital fandom, as well as for cross-disciplinary research in technology and the arts.
4. (Un)Creation: Theorizing fandom, industry, and praxis
[4.1] Scholars and fans are seeing the intersection of AI and fandom happen before their eyes, and new theoretical frameworks are needed in order to understand how these intersections —and exploitations —shape fan-industry relations. We do not mean to criticize individual fans who use AI but rather criticize the industries that put forth AI as an alternative to the creative process. We put forth the term (un)creation to describe this automation.
[4.2] In our view, automating creative processes via machine learning perverts that process. It is similar to the kind of repackaging Scott (2009, ¶ 1.1) describes when she writes, "Commodity culture begins selectively appropriating the gift economy's ethos for its own economic gain." Capitalism creates systems that pervert the relationship between human and technology, eliminating the praxis of creation from the final product, and positioning the end product as a commodified goal. Social science has noted the way platforms generate exploitive relationships with their users. Johanssen (2021, 112) articulates this process in the context of social media, but we may apply the kind of "data perversion" he discusses to fans and AI as well. As he writes, "users are simultaneously loved and abused, humanized and dehumanized, by platforms, or rather the developers and owners of them, today." In a system in which novel creative labor is bypassed, creative labor is devalued until it becomes a hollow representation, and we term this process and product uncreation. We can apply this relationship to fandom's creative culture, which simultaneously becomes the building blocks of generative AI, while being eroded by capitalistic systems that create generative AI tools and incentivize their use by fans. Whether end products created by AI hold legitimate creative value is, in our view, irrelevant here; our concern is with systems of power and their relationship to fan cultures.
[4.3] In line with this framework, we believe the technology industry poses a greater exploitative threat to fans and will continue to do so as AI technologies evolve. We call upon fan studies researchers to expand our frameworks of fan-industry exploitation to include problematic systems of data-driven capitalism. Explorations of fan-industry relationships have focused more closely on how media producers engage with fans and have, by comparison, neglected the technology industry as an additional exploitative force on fan cultures. Research foregrounding technology and media industries as twin exploitative forces, in which fans are simultaneously given influence and taken advantage of, can address these issues. Existing tech industry research focuses on censorship by platforms of fan expression that skewed too queer, feminine, or otherwise objectionable (a famous example being the LiveJournal Strikethrough). Kohnen (2017) notes that industry is wary of fans' expression, often shutting down fan wikis or websites, to use another example. We note here a need for scholarship that goes beyond censorship concerns and focuses on issues of data ownership, exploitation, and direct co-option of fandom via automating creative labor using existing fan works. As technology and fandom move forward in this new age, corporate big-tech developers will continue to catch fandom in the cross fire as they attempt to further automate creativity, placing fans in an ever more precarious place. The pursuit of profit, and of so-called easy alternatives to creativity, will continue to render fan work devalued and only useful to the extent that it generates capital.
[4.4] As fans, we value fan creations as whole, individual, novel objects grounded in love and joy. As researchers, we believe fan studies needs to place further emphasis on how data capitalism (Sadowski 2019) disadvantages fans. While fans' interactions with producers continue to be an important direction, we believe that, in the data-driven economies we live in, these twin exploitative forces cannot be separated. The pursuit of automated alternatives to the arts will continue to cause harm, particularly for fans. We have presented one framework for thinking about generative AI in relation to the arts and fandom, and we hope future fan-industry scholarship will continue to think through challenges posed by generative AI. As AI becomes an undeniable part of contemporary life and grows further entwined with the creative industries, it will be essential to center creatives (both fan and professional) in these discussions. We hope this article assists future fandom scholars in thinking through the intersection of fans and machines, and we hope this special section begins a fruitful research dialogue within this space.
5. Acknowledgments
[5.1] We would like to thank Hannah Judson and Carolyn Radillo for their comments and assistance on early drafts of this manuscript.