Composing with AI

Composing New Urban Futures with AI: Speculative Design and AI Text-to-Image Synthesis

Jamie Littlefield Texas Tech University

Introduction

In 2010, I co-founded a local organization dedicated to creating safer streets in Provo, Utah—a mid-sized city about an hour South of Salt Lake. The primary goal of our grassroots group was to encourage the construction of more walkable, bicycle-friendly, and accessible infrastructure in the built environment. As the organization matured and eventually became a 501(c)(3) non-profit, I often found myself debating developers and city officials in the public forum, in both textual exchanges (online forums, newspaper op-eds, mailers) and in face-to-face environments (town halls, engineering open houses, city council meetings). I soon discovered that citizen advocates are disadvantaged in debates about city infrastructure due to our lack of access to expensive, specialized technologies.

The professionals designing U.S. cities access industry software to extract place-based data and compose rhetorically persuasive visual renderings of the infrastructural futures they propose. Architects have AutoCAD and Revit. Urban planners rely on ArcGIS and CityEngine. City engineers use Infraworks and MicroStation. While these technologies are often described as helping professionals plan projects, they also work in tandem to help professionals communicate potentialities. Through polished renderings of street re-designs, skyscrapers, pedestrian crossings, and bus stops, members of the public are prompted to visualize the city not as it is but as it could be. City planning professionals are engaged in a perpetual act of storytelling in which “alternative versions of city space are explored, evaluated, and enacted over time” and “designers and architects envision new possibilities…then work to build them in physical form” (Hoffman, 2022, p. 11).

Without access to professional tools, citizens and grassroots advocates often struggle to convey alternative possibilities. For example, street safety organizations may have to pay outside consultants significant fees to create alternative images of sidewalk designs that prioritize vulnerable road users, such as people using wheelchairs or families with strollers. Visual rhetoricians have also attempted to demonstrate urban possibilities through free or low-cost design programs such as SketchUp, the method used by Gries et al. (2020) to counter gentrification by re-imagining a civic space at the University of Nevada, Reno. However, these tools can be time-consuming, and the results are generally more oriented toward communicating concrete design details than constructing rhetorically persuasive, stylized imagery.

Now, with the popularization of text-to-image tools, the disparity between professionals and grassroots communicators is beginning to decrease. AI is threatening to disrupt the monopoly urban professionals hold on a number of communicative affordances. Text-to-image synthesis (the processes used by programs such as DALL-E, Midjourney, and Stable Diffusion) is an emergent AI technology offering underresourced communicators a way to produce customized renderings with a lower price tag and a lower educational barrier to entry. A grassroots urbanist group might generate an image that adds a streetcar to an existing road, re-imagines an entire neighborhood around different values, or displays a visual prototype to demonstrate the consequences of a particular decision.

In my professional communications work, I’ve engaged with a number of non-profit leaders and community advocates who are beginning to explore how AI tools can help them more effectively communicate with the public. In 2022, social media saw an influx of AI-generated imagery showing cities not as they were but as they could be. That summer, I released a short prompt book for this niche audience, “Re-Imagining Urban Spaces with DALL-E AI,” which has been downloaded by just over 1,500 early adopters as of this writing. A second wave of AI-generated cities appeared on social media in late 2023, with the release of Microsoft Bing’s no-cost image generator powered by DALL-E 3 and ChatGPT. Social sites like X (formerly Twitter) circulated user-created images such as a Golden Gate Bridge favoring trolleys and pedestrians, a Baltimore Inner Harbor with extensively remodeled public spaces, and dozens of public beaches adjacent to muti-use paths rather than highways. For example, using DALL-E 2’s outpainting feature, I was able to take photos of real streets in my own community and edit them to include AI-generated elements such as highways, walking paths with native foliage, and water features (See Figure 2).

This chapter critically reflects on my experiences with AI text-to-image synthesis as a practitioner composing materials for grassroots advocacy and examines how this practical knowledge might translate to a classroom setting. I outline the essentials of how these tools work and how they offer new communicative possibilities to underresourced groups such as fledging non-profits and young political movements. I also describe some of the risks and challenges of AI text-to-image synthesis, including the possibility of value-locking the future. Finally, I suggest a method for critically engaging with AI text-to-image synthesis: speculative design.

Speculative design is a method used by designers, writers, and artists to envision and create concepts for possible futures rather than practical solutions for the present (Hoffman, 2022; Galloway & Caudwell, 2018; Dunne & Raby, 2013). It goes beyond traditional design by questioning societal norms and considering the broader impacts of technology and innovation on our lives. Speculative design involves crafting artifacts (such as AI-generated images) that demonstrate or visualize potential futures and encourage us to think critically about what could be (Opel & Rhodes, 2018; Wilkie et al., 2017). In their constitutive text, Dunne and Raby (2013) argue that speculative practices can function to generate greater elasticity for human futures: “...by speculating more, at all levels of society…reality will become more malleable and, although the future cannot be predicted, we can help set in place today factors that will increase the probability of more desirable futures happening…” (p. 6).

The tools of speculative design are uniquely equipped to help us critically consider the ways that AI is bound up in time, continually replicating the discourses of the past and artificially projecting limitations onto the ways humans will experience the future. In the case of urban communications, speculative design projects can help non-profits and grassroots organizations persuade the public by presenting images that deviate from the detrimental patterns of the past. In the writing classroom, speculative design practices can provide students with concrete approaches to analyzing and composing AI generated images that demonstrate awareness of the ways AI is entangled with the past.

How Text-to-Image Synthesizers Generate Content

In order to discuss critical approaches to AI, it may be helpful to review the processes text-to-image synthesizers use (See: Ramesh et al., 2021; OpenAI, 2021; Reed et al., 2016). With an eye toward balancing concision with clarity (and an acknowledgment that many details are not included here), this section presents a generalized snapshot of how OpenAI's DALL-E generates synthetic images. While the specifics of these processes differ between applications (programs like Midjourney use GANs or generative adversarial networks while DALL-E does not, for example), many features are similar.

As a broad overview, DALL-E generates synthetic images through a five-stage process:

1. DALL-E is trained on image/text pairs. DALL-E is trained on an extensive dataset containing pairs of text descriptions and corresponding images. This foundational training allows the model to learn the relationship between textual descriptions and their visual representations. Both the images and the text descriptors are broken into small chunks of meaning called "tokens."
2. A user inputs a prompt, which is processed as contextualized tokens. After a user submits a textual prompt describing the desired image, DALL-E breaks down the prompt into tokens and contextualizes these tokens. This means it derives a dense representation for each token that understands not only its standalone meaning but also its significance in relation to the entire prompt description. DALL-E analyzes those tokens based on its past exposure to the tokenized pairs of images and text within the training data.
3. DALL-E sequentially generates a synthetic image. With the contextually enriched textual representation as a foundation, DALL-E begins the image generation process. This is not done in one attempt; rather, it is a sequential process, similar to how ChatGPT generates text. DALL-E predicts parts of the image, often patches, in a sequence, ensuring that the evolving visual content aligns with the input description's intent. The initial output becomes input as the image completes.
4. DALL-E refines the synthetic image. After the initial synthesis, DALL-E leverages techniques like temperature sampling to refine and diversify the generated image. By adjusting the "temperature" or randomness parameter, DALL-E can produce a spectrum of outputs for a single prompt. This variability ensures that the user can select the most contextually relevant and visually cohesive image from the generated set of image options.
5. The synthetic image (eventually) becomes a part of a new data training set. Although companies have been reluctant to provide details of what output becomes input for AI systems, it is widely expected that a large number of synthetic images will become a part of future training data. This might happen from formal processes to directly capture and include AI output or through the publication of synthetic images on webpages that are later scraped to form new datasets.

As an introduction to this process, students may find it helpful to think of text-to-image data training sets as old photo albums in which every picture is accompanied by a caption. As it studies this album, DALL-E learns to create its own pictures when given new captions. By breaking down these captions into smaller pieces (like individual words), DALL-E generates new images by remixing visual and textual meanings from past data.

Working with the Past to Compose the Future

While the purpose of generative AI is to produce something new, the process itself is inextricably bound to the old. Generating a new image through AI is an act of engaging with textual history. In practice, Large Language Models (LLMs) like ChatGPT and text-to-image synthesis tools such as Midjourney are fundamentally rooted in the past, trained on vast datasets that encapsulate historical texts, images, and multimedia. This historical orientation allows AI systems to access a rich mosaic of human conversation and creativity, generating images that resonate with cultural, historical, stylistic, discursive, or artistic contexts. These tools have the potential to amplify the intelligence of previous generations, offering a synthesized form of collective human experience. They allow for an interaction with a broad swath of human history and culture, potentially making the past more readily accessible and interpretable to the present.

However, AI systems also present new risks: they may inadvertently perpetuate the prejudices and limitations of past data by parroting "encoded biases without understanding the significant harm of the language it produces" (Johnson, 2023, p. 170). Historical stereotypes, outdated norms, and factual inaccuracies can be unwittingly regurgitated, potentially reinforcing harmful social constructs or misleading users (Anderson, 2023). When exposed to extensive training sets, generative AI tools have been shown to exhibit bias favoring the people and subjects most prevalent in the data (Byrd, 2023; Getchell et al., 2022). The overall size of a dataset is not an indicator of its proportionality or inclusivity.

As it re-mixes meaning from the past, generative AI does not compose on a blank slate. Instead, it generates content overtop a virtual palimpsest—a layered canvas of data, code, and pre-existing human input. This palimpsest is not neutral; it carries the biases, assumptions, and limitations of its human developers in addition to the dataset it was trained on. Because of this, the output of AI is a complex interplay between machine learning and the past socio-cultural contexts from which its training data originates. Both humans and generative AI are engaged in a continual process of layering new information over old, negotiating between past influences and present intentions. Understanding the complexities of AI-generated content requires recognizing it as part of a broader communicative ecosystem, where human and machine are influenced by layers of pre-existing context.

AI's entrenchment in the past generates a sort of algorithmic resistance to change, creating synthetic output that is subject to "value lock." Bender et al. (2021) describe "value lock" as the way in which "the LM-reliant technology reifies older, less-inclusive understandings" (p. 614). Rapidly spreading social shifts in the way people discuss an issue—be it the discourse surrounding the Black Lives Matter movement, the #MeToo movement's focus on sexual harassment, or the global conversation on energy production—may not be fully captured by LLMs reliant on training data from years ago. Accelerated changes in public opinion, such as Western attitudes towards same-sex marriage between the late 1990s and the mid-2010s, may be attributed, in part, to what Ridolfo and DeVoss (2009) refer to as "rhetorical velocity." These significant social shifts can occur as texts are rapidly circulated, re-mixed, and re-circulated throughout a population.

Faced with a kairotic situation, a human communicator may choose to give more weight to a new concept in the creation of content, while an algorithm may favor the replication of patterns more widely represented in decades of past training materials. Rather than facilitating a change in discourse or even remaining neutral, AI systems may actively obstruct social change by generating content that is untouched by the ideological shifts spreading rapidly through human networks.

Essentially, when it generates synthetic content, AI re-assembles units of meaning drawn from the materials of the past. While the historical orientation of generative AI tools offers certain benefits, deliberate approaches are needed to address its limitations. At this critical moment, we might stop to ask ourselves: If new texts continue stochastically (randomly) parroting the meanings found in the past, what kind of futures will result? When generative AI is designed to shoulder the weight of history, how does it impact discourse and design? What interventional strategies can we use to change course now?

Speculative Design as a Future-Oriented Tool

I suggest speculative design as a method uniquely equipped to help compositionists and other communicators respond to the challenges of emergent AI technologies. Speculative design is a practice that focuses on exploring and illustrating potential futures rather than designing for immediate practical needs. Speculative design stimulates thought, discussion, and critical reflection about potential societal, cultural, ethical, and environmental futures. It challenges existing assumptions and conventions about the role that products or services play in our lives, fostering new ideas about possible alternatives (Hoffman, 2022; Wilkie et al., 2017; Dunne & Raby, 2013). Speculations may center around hopeful possibilities or dystopian consequences resulting from present choices (Jagoda, 2023). Speculation grows from "a sensibility concerned with resisting a future that presents itself as probable or plausible," and contends that "no matter how pervasive the impasse may be, [the probable future] can never exhaust the unrealized potential of the present" (Wilkie et al., 2017, p. 7).

Practicing speculative design often involves crafting scenarios, stories, technological prototypes, or artifacts—like the "photos" of future streets and cities I generated in my non-profit work. These creations help demonstrate potential futures, making them more material for public audiences. Essentially, speculative design encourages us to think critically about what could be. It is closely related to critical design, a term coined by designers Anthony Dunne and Fiona Raby (2013), authors of the text Speculative Everything: Design, Fiction, and Social Dreaming. Both critical design and speculative design use critical, maker-style methods to challenge the status quo and encourage people to consider their relationship with the designed world (Faris & Holmes, 2022).

Speculative design offers several benefits that make it a valuable approach in both academic and practitioner contexts. First, it fosters creative thinking by encouraging communicators (designers) and researchers to step outside conventional boundaries and explore alternative futures. This imaginative exercise can lead to innovative solutions that might not be discovered through traditional design methods. Speculative design often evokes what Johnson & Johnson (2020) refer to as "ambient imaginaries," the sensory details that "focus the attention of publics deciding which parts of the material environment to carry into the future" (p. 15). These sensory-based approaches help audiences form affective connections with an unlived future. Second, speculative design serves as a tool for social critique, allowing us to question existing systems and norms by presenting alternative scenarios. This can spark meaningful discussions about societal values, ethics, and directions for future development. The artifacts created by speculative design are not end products; instead, they are a means for facilitating the kind of participatory public processes that result in the usable, citizen-created data discussed by Moore and Elliott (2016).

Third, speculative design enhances interdisciplinary collaboration by integrating insights from fields such as technology, social sciences, and the humanities, potentially creating a richer, more nuanced expression of human intent. At its best, it invites broader participation in the design field rather than reinforcing disciplinary boundaries. Finally, speculative design acts as a catalyst for public engagement. By presenting provocative or visionary scenarios, this method invites the broader community to participate in shaping their own futures, a move towards democratizing the design process. Research suggests that when communities work together to imagine new futures, they develop social resilience—a necessary element of navigating challenge and change (Hoffman, 2022, p. 5).

While AI is entrenched in the meanings of the past, speculative design asks us to put our intentionality into projecting possible futures with the hope that doing so will serve as a method of reflecting on and rectifying present action. This chapter does not propose that all AI-generated projects should be speculative in nature. Rather, I am suggesting that human users may be able to mitigate the sociocultural consequences of generative AI by acquiring and implementing the critical skills that emerge from engaging with speculative design practices.

Three AI-Generated Examples of Speculative Design

To examine the potential uses of speculative design in text-to-image synthesis, I will consider three generative examples. Each of these small, exploratory projects turns a critical eye on the work I've conducted in the area of human-scale, eco-friendly city development as a non-profit communications consultant. These synthetic images are not intended to be polished pieces; instead, they are examples of tactical communicative interventions that can be rapidly created by underresourced groups.

Generative Example #1: Tracing the Projected Past

This first project demonstrates the ways that AI text-to-image synthesis re-creates the conventions of the past and provides suggestions for interventions based on speculative design. To begin this project, I examined how generative AI uses the cities of the past to create new images. Over 14 consecutive days, I entered the same four-word prompt in Midjourney's image-generating Discord server:

ideal neighborhood. photorealistic.

Out of the four resulting image options from each of the 14 initial generations, I consistently upgraded the first image (on the top left of the image results). I archived these synthetic images and analyzed the results, as shown in Video 1. By using the same prompt, I was able to develop a small corpus of synthetic images. Rather than analyzing images individually, examining a collection enabled me to begin identifying patterns.

Users may be tempted to assume that the resulting synthetic images would be widely varied, considering Midjourney's extensive training datasets. "Neighborhoods'' have long existed in a wide variety of countries, cultures, and decades. "Ideal" is similarly a vague descriptor that could apply to many subjective contexts. However, in this experiment, Midjourney recreated many of the same details for the prompt in different configurations. Its "new" output reflected patterns from particular subsets of the past.

After compiling these synthetic images, I examined the threads that connected them. I found that Midjourney was projecting a very specific interpretation of “ideal” and “neighborhood.” The corpus of synthetic images tended towards:

1. Suburbanism. All of the images showed houses in suburban settings. There are no mixed-use options (i.e., combinations of different types of building uses such as commercial spaces, apartments. or local restaurants). Only one image displayed an urban area in the background.
2. Americanism. Although the specific locale is unclear, the general aesthetics of the streets suggest that they are located in the United States (or possibly Canada). The architecture appears to be Victorian or Edwardian, a style popular in the U.S. from the late 1800s to the early 1900s and characterized by integrated details, varied rooflines, bay windows, and ornamental trims.
3. Car-centricity. 100% of the synthetic images include cars using the public space in the street or sidewalk. By contrast, only 7% of the images included humans using these public spaces. No alternative transportation methods, such as light rail, scooters, or buses were identified. Additionally, only 42% of the synthetic images included usable sidewalks on both sides of the street (or one side if only one side was pictured) that would be unobstructed to a person using a wheelchair or pushing a stroller.
4. Classism. All synthetic images featured large houses (generally two or three stories) in good condition with front porches and mature trees in what could be considered historic areas.
5. Racial Exclusivity and Redlining. While the images do not explicitly depict people, they evoke the aesthetics of redlined communities and mid-20th-century suburban idealism. The visual elements—large homes, manicured lawns, and detached housing—align with the historical American Dream imagery that was explicitly marketed to white, middle-class families in the post-WWII era. The federal government’s redlining policies and exclusionary zoning laws contributed to the development of these suburban enclaves, shaping neighborhood access based on race and economic status. AI’s reliance on past datasets means that it unintentionally perpetuates these historical patterns, reinforcing a limited and exclusionary vision of what an "ideal neighborhood" should be.

By drawing out the patterns in these images, we can start to identify how the AI text-to-image synthesizer is replicating a certain interpretation of the past. We can also begin to imagine how using these synthesized images in communicative contexts could produce value-lock by limiting the way we think about the future of neighborhoods. From the position of my own grassroots advocacy, I am concerned that using these images might contribute to the replication of the same type of exclusionary spaces that are already present in the built environment.

Applying a speculative design approach can help users imagine alternative ways to think about the same concept. Speculative design asks us to consider what alternative futures might be available to us. What if we were to create and circulate images of ideal neighborhoods that were not suburban, American, car-centric, and upper-class? What other types of neighborhoods might be considered "ideal" in the future? Speculative design practices can help users imagine built environments that are less predictable and less inclined to perpetuate value-lock.

One approach to generating more speculative synthetic images is to use AI as a tool for what Deleuze and Guattari (2013) call "rhizomatic thinking." Rhizomatic thinking is a non-hierarchical, decentralized cognitive approach that allows for multiple entry and exit points, much like the underground stems of rhizomatic vegetation (like bamboo, ginger, or mint). Unlike traditional, linear ways of thinking that seek a single "correct" solution or approach to an answer, rhizomatic thinking embraces complexity, multiplicity, and interconnectedness. In the context of AI-generated images, this means moving beyond mere replication or simulation of existing visual forms derived from the past. Instead, AI can be prompted to explore a larger array of possibilities, each branching off from the other, to create images that are not just novel but also speculative. These images can challenge our preconceived notions of reality, aesthetics, and even the limitations of human imagination, opening up new avenues for inquiry.

After noticing the patterns that emerged in the first corpus of generated images, I worked to create prompts that challenged text-to-image synthesizers to connect the concept of "neighborhood" to concepts that might be considered more distant and less linear. I experimented with new nodes like "Afrofuturism," "design for women," and "agricultural urbanism." The key phrases used in these prompts were intended to serve as nodes for rhizomatic thinking, asking AI to tap into knowledge systems that were further from its original concept of "neighborhoods."

The resulting images (Figure 6) showed visual evidence of meaning drawn from alternative branches of thinking. In these images, we see: neighborhoods being used by many different people, bicyclists and pedestrians, shared public spaces, waterways, communal playgrounds, and different housing types. The synthesizers are still drawing from the past, but the generated content shows greater novelty because it is working with additional units of meaning. It is being asked to create connections between somewhat disparate pasts, between branches of thought that are less related. AI text-to-image synthesis allows users to intentionally and rapidly experiment with this style of rhizomatic thinking by generating content from multiple "nodes" through textual input, providing new possibilities for projecting alternative futures.

While connecting to alternative nodes through rhizomatic thinking helps generate novel combinations, each new conceptual branch carries its own set of value-locked patterns that deserve critical examination. When connecting to "Afrofuturism" as a node, the AI draws from existing visual conventions that often frame Black futures through a lens of technological dystopia and architectural monumentalism, replicating patterns from science fiction media rather than the full spectrum of Afrofuturist thought. The "design for women" node activates deeply embedded visual patterns from historically gendered spaces and marketing, defaulting to stereotypical feminine coding through color and ornamental choices that reflect decades of gender-segregated design practices. Agricultural urbanism visualizations tend to reproduce the aesthetic conventions of contemporary "eco-luxury" developments, pulling from marketing imagery that presents urban farming through a lens of white, middle-class lifestyle rather than connecting to the diverse cultural histories of community agriculture and food sovereignty movements.

These patterns reveal how, even when generating novelty through rhizomatic connections, AI systems reconstruct value-locked elements from each conceptual branch they access. This suggests that speculative design practices must involve not just making new connections between nodes but must critically examine what historical patterns each node brings with it. Practitioners can use this awareness to deliberately identify and challenge these inherited conventions—perhaps by introducing additional nodes that specifically counteract recognized patterns, or by crafting prompts that explicitly reject certain visual tropes. The goal is not only to generate novel combinations, but to thoughtfully curate which elements from each knowledge branch we want to carry forward into our speculative futures.

Composition instructors can guide students through rhizomatic prompt creation by teaching them to strategically combine conceptual nodes (see Table 1). Like assembling blocks of meaning, students can start with a Base Spatial Concept (a "neighborhood" or "public space"), then thoughtfully layer on Adjective Modifiers that shape its character ("sustainable," "vibrant"). By adding a Style or Movement node ("solarpunk," "indigenous architecture"), students begin making more distant conceptual connections. Additional Features ("green rooftops," "community spaces"), Environmental Context ("coastal," "urban forest"), and Social Dynamics ("intergenerational living") can then be mixed in deliberately, with students examining how each new node might introduce its own historical patterns or biases. For example, rather than simply accepting "Design a vibrant neighborhood with indigenous-inspired architecture featuring rooftop gardens in a mountain terrain for intergenerational living," students should interrogate what assumptions about class, culture, and accessibility each element brings. This process teaches critical awareness of how different knowledge branches interact while encouraging experimentation with unconventional combinations that might challenge value-locked futures.

Generative Example #2: Speculative Design as a Tool for Value Negotiation and Identity Formation

As a practice, speculative design challenges communicators to tell stories that prompt discussion about who we are and what we value. By crafting narratives that exist beyond the present moment, communicators invite audiences to confront their own assumptions, biases, and ethical frameworks. This form of storytelling catalyzes dialogue, encouraging people to negotiate their values and reconsider their identities in the face of hypothetical yet plausible situations. In doing so, speculative design transcends aesthetic or functional considerations; it becomes a tool for social critique and identity exploration (Snow et al., 2021).

A place can be understood as a space that is filled with identity, culture, and shared references, bound together in time. Non-places, on the other hand, tend to promote the solitary identity of individuals disconnected from the people immediately around them (Varnelis & Friedberg, 2008). Examining the way places are connected to human values can help us identify what Rai (2016) refers to as "the competing rhetorical frames that circulate within and are tied to literal places" (p. 34).

For this example, I used DALL-E 3 to generate images that promote discussion about shared values around vacant property within my community. This project was done in response to local debate around the need for more convenient parking versus the need for more public spaces.

To generate the first synthetic image, I attempted to create distance between the subject and the anticipated audience. I chose the Eiffel Tower due to its status as a cultural symbol and its physical and ideological distance from Provo, Utah (i.e., it is known and loved but is distant and not considered "ours"). I used DALL-E 3 to generate a synthetic image, removing the public space that picnickers and tourists enjoy below the tower and replacing it with a parking lot for electric vehicles.

For the second synthetic image, I attempted to close the gap between the place pictured and the anticipated audience. Using an iPhone, I took a simple photograph of an empty lot in a downtown Provo neighborhood. I then used DALL-E's outpainting tool to generate three images representing the materialization of different values: the construction of a parking lot, a bicycle-friendly community garden, and a neighborhood restaurant (Figure 8).

The purpose of these generations is not to needlessly provoke but to encourage community discussion. In these examples, I attempted to elicit responses that called for a deeper reflection on the ways public space is used and how that use reflects community values. Speculative design can draw people together around new kinds of inquiry: What do we gain when we use publicly owned space for parking? What do we lose? What does that say about us and our values? Are there specific circumstances where we might choose to trade the convenience of parking for another value, and how do we identify those circumstances?

Generative Example #3: Speculative Design & The Construction of Publics

This final example demonstrates how speculative design has the potential to create publics—groups of strangers who are united around a common text (written, visual, or multimedia). Drawing on the work of Dewey (1927), Habermas (1989), and Latour & Weibel (2005), writing researchers have long been interested in the ways that the creation of a text can construct a group of people and things empowered to take conjoint action (Gries, 2019; Moore, 2023, Preston, 2015; Boyle & Rivers, 2018). Such texts often prompt action through their circulation as a part of a larger rhetorical ecology, generating force through their association with other texts, rhetors, readers, histories, and materials (Rivers & Weber, 2011; Edbauer, 2005). By assembling a public with intentionality, communicators invite readers to see themselves and their involvement in an issue in a particular way and, potentially, work with each other to change the course of the future (Warner, 2002; Rice, 2012).

For this project, I wanted to communicate the urgency of responding to the environmental threats facing Utah due to climate change and highlight the need for alternative transportation methods like infrastructure that supports walking and bicycling. I chose to focus on the Great Salt Lake, a rapidly declining body of water receding due to recent droughts. For generations, the lakebed was used to dispose of toxic chemicals, which now threaten to become airborne as dust storms. I used Midjourney to generate a synthetic image displaying a dystopian possibility for the future of Salt Lake—a dried, cracked lakebed releasing toxins toward the city (Figure 9).

Following the practices of speculative design, I generated a companion image to suggest potential technologies that might be needed to mitigate this dystopian future (Figure 10). Refined in Canva, the image displays a prototype of a colorful mask combined with an AI-enabled air quality sensor that could be distributed to Utah children in areas experiencing toxic dust exposure from the declining lake. While K-12 students in our area are already required to stay indoors during recess on days with dangerous particulate matter readings, such a device could allow children access to outdoor spaces if air quality continues to decrease. These corresponding images invite readers to consider how the actions taken now may impact the nature of our community and the ways we exist within it in the future.

When combined with narrative, speculative imagery can serve as a rhetorical tool to gather the people and materials needed to enact change. In the case above, the images are intended to evoke emotion around an undesirable future and its technological implications and, potentially, generate the affective energy required to initiate public action. Here, again, the purpose is not to needlessly provoke but to identify ways that AI text-to-image synthesis can help us speculate about the future in order to change the present. In this case, generative AI is a valuable tool for rapidly creating narrative-based images and prototypes. It allows users to experiment with variations, pulling together materials of the past to imagine solutions for problems in the future. Most importantly, it offers new tools for communicators to pull together groups of actants around shared concerns.

In my non-profit work, the theories behind the construction of publics have proven particularly relevant. Our task is to assemble groups of people and things (publics) who are inspired to work together to reassemble the material elements of the built environment (streets, bike lanes, parks, etc.). Our recent air quality failures might be viewed as a breakdown in our attempts to effectively form publics—we have not yet been able to assemble the kinds of groups that are capable and motivated to address the environmental problems we face. By providing communicators the chance to rapidly experiment with different types of content creation, we may be able to more effectively assemble the types of publics we need.

It is important to note here that emergent technology may end up impacting not only what publics are formed but how publicness itself is experienced. That is, technology has the potential not only to change the membership of a public but to change the way people/things are drawn into a public and the way they experience that association itself. In Composing Place: Digital Rhetorics for a Mobile World, Greene (2023) points to the ways that technology may be used to create new kinds of public orientations:

Rather than reducing the function of a digital text to its ability (or inability) to persuade a pre-given group of people, we should also attempt to understand how emerging technological infrastructures facilitate new formations of public-ness itself. (p. 40).

As technologies like augmented reality (AR) gain traction, humans may begin experiencing material places (like bus stops or public plazas) through interfaces that restrict or emphasize particular aspects of their experience with the built environment. Wearing AR headsets, one user may see factoids about local art in his visual field, while another may see evidence of local homelessness as a nondescript blur. Such applications function to "not only orient our physical position within a space but also our rhetorical position" (Crider et al., 2020, p. 9). Speculative design offers avenues for drawing users away from their individualized AR experiences and toward shared, material ways of addressing community challenges.

Conclusion

This chapter has considered how speculative design can serve as a tool for the critical examination and use of AI text-to-image generation. Although the focus has been on visual texts, a number of the concepts are transferable to other modes of communication. In many ways, the emergence of technologies reliant on large language models is an opportunity for writers to think more deeply about their own craft. As Walter Ong (2012) resonantly explained, writing itself is a technology. The act of writing has always existed in a space of tension between the past and the future. Through the manipulation of material tools and human symbols, we generate texts that encourage new understandings by remixing past understandings.

Beyond writing itself, many of the tools professionals use to gather and communicate data are also entangled with the past. In the field of urbanism, professional tools used by city planners and engineers are explicitly designed to recreate (rather than re-imagine) the existing elements of the built environment. It seems only sensible to design new streets based on our knowledge of past streets rather than rethink the nature of the street itself each time a road is called for. However, this way of engaging with the world creates entrenched patterns that can replicate bias and exclusion (like sidewalks without wheelchair access or public spaces only accessible to car owners).

Emergent technologies do not fundamentally change the purpose of communication. However, they do pose new questions, especially as AI begins to act as a mediating influence between the communicative acts of the past and the communicative acts of the future. Generative AI technology allows us to shift some of the labor of symbolic work to algorithmic functions. But, by doing so, we may risk offloading the work that is most important: our ability to imagine and communicate new futures—possibilities that loosen our entanglements to the past and propel humanity forward. Speculative design can help human communicators re-consider not only what kind of work they do but also what kind of beings they are. Specifically, in Gries' (2019) terms, it can help us remember that we are "assembling beings" with the "ability to assemble and distribute discourse that can, in turn, assemble and reassemble bodies around a shared concern" (p. 336).

A speculative approach to generative AI can help us further examine what it means to exist as an assembling being and how new things (especially new futures) emerge from the fragments of the old. Such a perspective can help us identify new ways to work with AI systems serving as mediators between ourselves and a collective, partial past.

References