‘Someone who uses it better’: Speculative fiction as method for AI work refusal in cultural geography

Abstract

Work practices and workplaces are central to discourses about artificial intelligence (AI). This centrality is reflected in the ongoing introduction of AI into existing work processes and popular predictions about how AI will lead to job replacement in the future of work. We present a short speculative narrative that mediates between dystopian and utopian depictions of AI and its refusal to loosen dominant tropes that characterize AI’s uptake in workplaces. We suggest that narrative can be a method or practice for doing cultural geography, and our narrative offers a glimpse into our protagonist’s working life in an AI-dominated future and her more and less subtle acts of refusal therein. Informed by critical research in feminist economic geography and beyond on the future of work, our narrative dramatizes, in hopefully unexpected ways, what an AI future of work (refusal) looks like.

Speculative fiction as refusal in cultural geographies

Artificial intelligence (AI) is central to debates on (the future of) work as a technology poised to inevitably disrupt workplaces and put many out of work. ¹ Despite the launch of browser-based, consumer-facing large language models like ChatGPT 4.0 in 2022, AI’s actual impact on job loss remains, to some extent, speculative, exaggerated, and hype-driven, based on predictions and forecasting, rather than on empirical research. ² Yet, AI is having a ‘disruptive’ impact on many jobs, including the journalistic work depicted in our narrative below. This impact is more complex than simple replacement and stands to devalue human labor and integrate AI within existing workflows. ³

Following epistemological trends central to cultural geography that critique the production of facts as outputs of objective science, ⁴ we position fiction as a creative method for doing cultural geography that allows us to explore the interface between fact and fiction, reality and imagination. ⁵ We present a short work of speculative fiction that engages with the theme of AI work refusal – how we might reject, transform, or subvert AI’s encroachment on jobs and work processes ⁶ – in at least two ways. First, we use narrative-as-method as a refusal of the faux-scientific discourses that predict clear, one-directional, and inevitable links between technological change and job loss. Second, our narrative depicts refusal in practice: our protagonist refuses AI in her regimes of production and social reproduction in complex ways. Our narrative is, thus, a method to creatively practice refusal by moving ‘in-between “real” and “imagined” worlds’ and teasing out what refusal looks like in a future working life infused with AI. ⁷

Our narrative is informed by critical research in feminist economic geography on the future of work that conceptualizes ‘workplace’ in an expanded sense, beyond the single worksite (e.g. an office) and as encompassing home-regimes, other socially reproductive spaces, and the transitional space of the commute. ⁸ Other tenets of feminist economic geography central to our narrative include theorizations of social difference and inequality ⁹ and the idea that colonial-capitalism will always demand more and harder, rather than less and better work. ¹⁰ In our narrative, the future of work is not one in which humans have been replaced by AI. Instead, AI devalues human labor, as the cost of producing, storing, insuring, and maintaining technology, computers, and robots always exceeds the cost of paying workers less than the value of their labor time. ¹¹ It is a future where AI hasn’t become substantially more sophisticated than it is today and where AI remains rife with glitches, biases, and unintended consequences. ¹² Our depiction reflects a conviction that AI’s function is to extract surplus value from already poorly remunerated workers, rather than to ‘solve’ technological problems.

Following Bonnie Honig in A Feminist Theory of Refusal, our narrative roughly charts refusal across Honig’s three domains of inoperativity (exposing a politics of utility), inclination (examining care and its attendant violences), and fabulation (creatively writing beyond the knowable archive). ¹³ Our protagonist, Vega, ¹⁴ subverts (but, because of her continued dependence on the market for social reproduction, cannot reject ¹⁵ ) her use of AI as a writer, pushing back on her role as a closely monitored AI-enhanced propagandist to support her coworkers and progressive political insurgents beyond her workplace. Her actions create a horizon of future action that is heterotopic ¹⁶ and speculative: a new world to come. ¹⁷

Honig’s framework provides a complement to another way of theorizing refusal in labor and economic geography: Katz’s resilience, reworking, and resistance. ¹⁸ Honig’s inoperativity, inclination, and fabulation subvert the oft-assumed hierarchy of Cindi Katz’s three Rs and productively unsettle this genre of refusal. What might be seen as ‘only’ resilience is instead a form of expectant waiting ¹⁹ for change that is being actively brought about, not just a making do while being unable to change one’s immediate circumstances. In our story, resilience is a condition of reworking and resisting in ways that render both concepts interrelated in an extended ‘arc of refusal’. ²⁰ Further, harkening back to myth through allusions to Honig’s interpretation of Euripides offers a partial antidote to notions of the future as only linear, sequential, and progressive. ²¹ Honig’s text, and our narrative, thus loosens ²² notions of refusal in geography, beneficial to cultural geography because of its interest in genres of refusal beyond the conventional and academic. ²³ The narrative, in form and content, raises questions about what the futures of AI work (refusal) might look like and how conceptualizing these futures differently (i.e. through fiction and other creative methods) is valuable.

Someone who uses it better

Vega waits for the bus that will take her to the office, her thoughts wavering between her previous workday, her apartment shared with five roommates, and her upcoming commute. She glances at her phone and scrolls through notifications from applications assigned by her phone’s AI-enhanced operating system. Vega scrolls past NewsApp notifications: ‘City Defends Against Foreign Insurgents’ and ‘Police Closer to Tracking Down Exiles’. The headlines, as usual, are contradictory. AI-generated, human-edited nonsense, as Vega knows well: writing these headlines is her job. Those now living outside the city left so long ago as to be hardly remembered. It’s harder to leave the city now, but rumours persist: autonomous communities, groups of women, femmes, queers, and yes, some men too, all rejecting – or rejected by – the AI revolution, beyond the city, waiting to return one day.

A notification from her roommates discussing rent increases and a notification from RenterAI remind her that rent is due. ²⁴ Opening her FinApp, Vega is greeted by an AI financial assistant sending advertisements for credit-, crypto-, indentured, and deferred payment products. Before she can pay her rent, the AI assistant triggers a 45-second ‘engagement period’ (the cost of ‘free’ online banking), demanding that Vega read through financial tips and answer questions. Resigned to the necessity of performing her financial literacy and providing data to the application, Vega taps through the options. ²⁵ The transaction complete, her rent is paid.

Vega boards the bus for her commute to Writers Block, clusters of office towers that investor-targeted ads boast as where ‘the world’s best AI and non-AI certified content is produced’. From the bus window she sees nuclear towers that power the AI-enabled city peppering the skyline. Cooling stations betray the vast server farms (and army of underground workers managing and maintaining them) that lie beneath. ²⁶ Thankful that she works above ground, Vega pulls her laptop from her bag. An advertisement for a certification course to learn about the newest AI models appears on her screen: ‘Remember, you won’t be replaced by AI, you will just be replaced by someone who uses it better’. ²⁷ Palms sweaty, she considers next week’s rent payment and recalls that receiving work for the day is never guaranteed. Her cursor hovers over the ad.

Before she can click, a series of inaudible vibrations come from her bag, and Vega remembers that she has another job to do. Alert, she looks out of the window: she’s nearly at the assigned location. Glancing around at the other passengers and toward the bus’s surveillance system, she leans toward the bus’s external wall and pulls from her bag a last-century device, retrofitted to refuse connection to the city-network and to connect, instead, to now-illegal 5G networks. Her connection is secure, and Vega pastes a pre-written message from her clipboard and taps send. The network vanishes; Vega slips the device back in bag. She feels hot; the clatter of the bus and the cacophony of the city press up against her, seeming too loud. Trembling, she stands up ²⁸ and turns toward the other passengers, sure that someone will be staring at her. But her fellow passengers, phones in hand, are swiping and scrolling, reacting to AI-generated games, notifications, and advertisements. Vega breathes.

The bus stops, and Vega walks toward one of the office towers. She passes through a turnstile to receive her work assignment. A printed chit – the digital ones were too easy to fabricate, including by writers like Vega – shows two numbers. The first is Vega’s call number. If she’s unlucky, her number won’t be called for several hours, if at all, cutting her earning capacity for the day. The second number ranks her last workday’s performance based on the number of completed assignments, typing speed, message response time, and more. Perennially annoyed by management’s attempts to pit her against her co-workers, Vega’s eyes flick to the ceiling. The number fails to capture ‘performance’, since in practice Vega trades assignments back and forth with her colleagues based on their individual preferences. They make jokes, keep one another going, check one another’s work, and help newer colleagues navigate interactions with the AI management systems. ²⁹

Like yesterday, Vega’s work number is called promptly. Used to the rhythms of the AI management algorithm, she knows that receiving the same assignment for several consecutive working days signals that something is amiss. She takes the elevator to the 60th floor to receive her assignment. Doors open to lush offices where highly paid union-certified ³⁰ RealWriters type their mandated 100-words-per-minute, every keystroke recorded on their personal blockchain to create a pristine record of human-generated writing. Now that 99% of text is AI generated, the blockchain is essential in the authentication of human-generated text. ³¹ NewsApps like Vega’s place a premium not only on the scarce ‘real’ human word but also on its authenticatability. While AI-written articles still form the largest part of her NewsApp’s business model, AI systems crave new inputs, having long ago consumed the archive of human-generated writing.

In contrast to the highly valued, human-authenticated content of RealWriters, AI-enabled writers like Vega are editors, assemblers, rearrangers, stitchers, fixers, and generators, in a multiplication of writing-adjacent, AI-enhanced roles. ³² AI doesn’t reduce the need for human labor but keeps its costs low, subordinating it to an ecosystem of bespoke and open access, centralized and forked, AI systems. Humans produce generic copy to be eaten up by NewsApps’ proprietary AIs, edit AI-generated text to create ‘unique’ content, stitch together outputs from AI models to produce better (or, at least, different) results, and fix errors and glitches with outputs as they emerge. NewsApps have unique AIs for different kinds of reporting, audiences, and issues and employ back-end programmers who tinker with AI models to achieve the ‘right’ tone (‘“right” determined by whom?’ Vega always wonders) across content verticals.

Vega is only on this floor to hand in her work number and receive her assignment: she will return to a cubicle on the 12th floor, where she’ll stay as late as is necessary to complete her piecework for the day. ³³ She hands in her number and, as she expects, receives the same assignment as she did yesterday. Smiling to herself, she now knows that the algorithm has been compromised, although she cannot be sure by whom. Vega, glancing at the RealWriters before the doors close again, takes the elevator down to where she’ll begin her workday. Today, she will be working with FederalistApp14.2, an AI model that generates predictive content for ‘serious’ readers of city politics. 14.2 is an improvement on the previous version: it is less likely to generate nationalist screed, but Vega’s job isn’t always to eliminate the racism embedded in the models. It is an election year after all. ³⁴

Vega will review and revise AI-generated articles according to the guidelines she’s received. She knows that she’s doing the work of a propagandist, but also that most of her readers see through this façade. But for the second time today, Vega has another job: she’ll also be removing, editing, and manipulating the fear-mongering messaging in articles about the communities outside the city, rewriting the archive to create sympathy for those who have left and to draw attention to the city’s persecution of them. ³⁵ She’ll play a small part in making the city ready for their return. Her mind flits to the last-century device hidden in her bag. She opens her first assignment.

AI work refusal, or, thinking forward with speculative fiction

Our narrative offers a perspective on AI work refusal foregrounding its potential, messy, imperfect, and unfolding futures. ³⁶ Refusal is present in the story, sometimes only obliquely, through Honig’s concepts of inoperativity, inclination, and fabulation and through Katz’s resilience, reworking, and resistance. In reference to Honig, a politics of inoperativity inflects the whole story, dramatizing the uses of human versus mechanic labor. On her commute, Vega, inclined against the exterior wall of the bus, communicates illegally with someone, implicitly the group who have left the city, demonstrating care for these political assailants. The last paragraph shows Vega manipulating the stories she works on in an act of fabulation, to cultivate support for those who have left. Vega is resilient in her coping with low pay and difficult working circumstances. She reworks her and her colleagues’ circumstances by exchanging tasks and informing them about workplace politics. Turning to Katz, Vega demonstrates resistance through her illegal communications and manipulation of the stories she works on. While these acts of AI work refusal are significant, they do not mean that Vega is able to wholly transform her circumstances: Vega still feels anxiety about being replaced by AI or by someone who uses it better, she submits to her FinApp’s gamified advertising regime, and, of course, she continues to work, remaining structurally dependent on the market for her social reproduction.

Our narrative captures some of the messiness and complexity of embodied capacities for (in)action across AI-saturated regimes of production and social reproduction. In connection with critical literature from feminist economic geographers and others on the future of work cited prior to the story, Vega’s narrative helps us to question received wisdom and economistic narratives about the future of (AI) work and its refusals. Using narrative draws attention to how research about the future of work also involves fabrication, albeit fabrication that is allowed to be labelled as ‘science,’ despite, often, the absence of real empirical evidence to back up its claims about immanent and widespread joblessness. Speculative fiction, as a practice in cultural geography, allows us to question these boundaries between fact and fiction. In both form and content, therefore, (our) narrative acts as a mechanism of refusal. This is not only a refusal of AI and work, but also a refusal of the conventional and dominant frameworks, narratives, and genres used to understand the relationship between AI and work in the future of work.