Content uploaded by Nataliia V. Laba
Author content
All content in this area was uploaded by Nataliia V. Laba on Jun 12, 2024
Content may be subject to copyright.
https://doi.org/10.1177/01634437241259950
Media, Culture & Society
1 –22
© The Author(s) 2024
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/01634437241259950
journals.sagepub.com/home/mcs
Engine for the imagination?
Visual generative media and
the issue of representation
Nataliia Laba
University of New South Wales, Australia
Abstract
Visual generative media represent a novel technology with the potential to mediate
public perceptions of political events, conflicts, and wars. Seeking to understand a
visual culture in which algorithms become integrated into human processes of memory
mediatization, this study addresses representation in AI-generated war imagery. It frames
AI image generation as a socio-technical practice at the nexus of humans, machines,
and visual culture, challenging Silicon Valley’s prevailing narrative of visual AI as “an
engine for the imagination.” Through a case study of AI images generated in response
to verbal prompts about Russia’s war against Ukraine, I examine the representational
capabilities and limitations of the text-to-image generator Midjourney. The findings
suggest homogeneity of visual themes that foreground destruction and fighters, while
overlooking broader contextual and cultural aspects of the Russia-Ukraine war, thus
generalizing the depiction of this war to that of any war. This study advances the research
agenda on critical machine vision as a transdisciplinary challenge situated at the interface
of media and cultural studies, computer science, and discourse-analytic approaches to
visual communication.
Keywords
AI-generated images, artificial intelligence, image generators, Midjourney,
representation, Russia-Ukraine war, visual culture, visual generative media
Latour (1999): Do you believe in reality?
ChatGPT (2023): I’m just a computer program, so I don’t have beliefs or consciousness. I
process and generate text based on patterns and information in the data I was trained on.
Corresponding author:
Nataliia Laba, School of the Arts and Media, University of New South Wales, High Street, Kensington,
NSW 2052, Australia.
Email: n.laba@rug.nl
1259950MCS0010.1177/01634437241259950Media, Culture & SocietyLaba
research-article2024
Main Article
2 Media, Culture & Society 00(0)
This banality of machine learning is also its power. It’s a generalised numerical compression of
questions that matter (McQuillan, 2018).
The art of the past no longer exists as it once did. Its authority is lost. In its place there is
language of images. What matters now is who uses that language for what purpose (Berger,
2008).
AI-enabled image generators have captured the zeitgeist in the past year, marking a
major shift in the landscape of visual communication. This shift began with OpenAI’s
breakthrough with DALL·E in January 2021. DALL·E, a linguistic fusion of Salvador
Dali and Pixar’s WALL-E, initiated a surge in visual generative media such as
Midjourney, Stable Diffusion, Craiyon, and others emerging from labs of private
commercial entities and prestigious research institutions. While many produce and
share AI-generated images on social media platforms, others engage in spectatorship
of the captivating capabilities of AI. Popular media, particularly Wired and magazines
like Newsweek, amplifies this fascination, reporting on the advances of text-to-image
generation technologies (e.g. Christie’s, 2018; Kelly, 2022) and speculating about
their promising future (e.g. Marber, 2023).
The technophilic AI rhetoric echoes the Californian ideology (cf. Barbrook and
Cameron, 1996) – the one which provides a contrast to the real world against the ideal
vision of the future, framing technological advancement as unavoidable but also as
something yet to be achieved. As Forlano (2021) observes, discounting the present for
the future helps technology companies and data evangelists conceive of the future as
either a solvent to existing social problems or as a tool that will amplify human abilities.
In this way, commercial actors conveniently “absolve themselves of responsibilities for
the present” (Dourish and Bell, 2011: 22) by placing emphasis on innovation, efficiency,
and resolution of societal challenges through the application of cutting-edge technolo-
gies. This discursive construction of technology becomes a powerful instrument for
charismatic digerati – intellectuals advocating for generative AI – sharing their techno-
utopian rhetoric with the public (see e.g. Vincent, 2022). However, such framing obfus-
cates a significant detail: when people engage in AI image generation, they participate in
a social action that is highly homogenized, regimented, and proceduralized.
Throughout much of the history of digital technologies, a prevailing assumption,
commonly referred to as “the Lovelace objection,” permeated computer science cir-
cles. This notion posited that machines cannot create anything original but only do
what they were instructed to do (Abramson, 2008; Gray and Suri, 2019). Now, perhaps
for the first time in the human history, a technological tool has transcended its tradi-
tional role as a medium, evolving beyond the likes of historical innovations such as the
printing press, television, or telephony and producing visual cultural artifacts “on its
own.” The current iteration of visual generative models has reached a level that sur-
passes the average human ability to produce and manipulate images. These models not
only generate original visual artifacts (in the computer vision sense) but also outper-
form previous state-of-the-art methods of image production. While AI-enabled image
generators are reportedly used “for fun” at present (Wilson, 2023), it is important to
examine their potential trajectory, particularly because there is a looming prospect for
Laba 3
AI images to mediate public perceptions of real-world events, including political
developments, conflicts, and wars.
This paper has two research objectives. Firstly, I conceptualize AI image generation
as a socio-technical practice at the nexus of humans, machines, and visual culture, chal-
lenging the dominant commercial narratives of visual AI as a tool for enhancing artistic
imagination and visual production skills of human actors. Such an understanding is nec-
essary in the times of accelerated neoliberalism (Tulloch and Johnson, 2021: 923) – an
exponential outsourcing of individual choice to algorithms, intelligent assistants, apps,
gamed environments, and – in the case of visual generative media – the computational
power of deep neural networks. Secondly, through a case study of AI images generated
in response to verbal prompts about Russia’s war against Ukraine, I investigate represen-
tational capabilities and limitations of the text-to-image generator Midjourney and dis-
cuss implications of visual generative media for memorialization of conflicts and wars
through what Bucher (2017) terms algorithmic imaginary, or (human) “capacity to
imagine, perceive and experience algorithms and what these imaginations make possi-
ble” (p. 31). Because the impact of generative media on visual culture cannot be produc-
tively studied without addressing algorithmic and machine learning processes, I adopt an
analytical focus that encompasses both the social and the technical.
The paper is structured in a way that responds to these two objectives. In the follow-
ing section, I consider the notion of “machine vision” from the perspective of media and
cultural studies, zooming in on several issues surrounding human-machine interaction in
the context of mediatization of war, before introducing research questions and the con-
ceptual framework. Next, I address the technical aspect of AI image generation in a way
that is accessible to digital media and communication scholars, outline the theoretical
framework, and explain how it informs the selection of methods. Lastly, I discuss find-
ings, limitations, and future research directions, focusing on human-machine agency and
representational problematics of visual generative media.
Machine vision and memory mediatization
Proliferation of digital technologies has impacted, among many social and cultural prac-
tices, the ways in which individuals and communities remember and forget, providing
new opportunities to formulate, reinforce, or challenge interpretations of the past
(González-Aguilar and Makhortykh, 2022). Mediatized memory is the one which “flows
across territorial and social boundaries” (O’Connor, 2022: 635) interwoven with arrange-
ments of cultures, agents, and materialities of production technologies (Lohmeier and
Pentzold, 2014). When applied to the context of conflicts and wars, collective memory
has traditionally been impacted by conventional media, particular by news photography
(Griffin, 2004; Parry, 2010). Human-produced war photography is appreciated as a con-
tinuing practice of cultural production within specific historical, cultural, and political
circumstances (Parry, 2011), and, as Griffin (2010) argues, a full consideration of any
image of war must include an analysis of these conditions. John Berger might also say
that war imagery is implicated by how the subject has been seen by other people and that
“the specific vision of the image-maker was also recognized as part of the record”
(Berger, 2008: 10).
4 Media, Culture & Society 00(0)
But what is seen in AI-generated war images, and what remains unseen? Generative
models bring about changes in agentive nature of technology (Natale and Guzman, 2022)
in that AI is increasingly seen as an entity operating in ways too complex for human com-
prehension, with limited opportunities for human actors to tap into system’s infrastructure.
Due to algorithmic opacity of visual generative media, there exists a perpetual tension
between the perceptual bias arising from inductive biases in a machine vision system that
determine its capability to represent the world (Offert and Bell, 2020: 1333) and the
assumed “ways of seeing” – a diverse range of perspectives and realities that the system
lacks the capacity to represent. As a result, AI-generated images can reinforce dominant
hegemonic narratives in representing collective memory (Makhortykh et al., 2023).
Hegemonic AI can act as a “bio-necro-technopolitical machine that serves to maintain the
capitalist, colonialist and patriarchal order of the world” (Ricaurte, 2022: 727).
Thus, machine vision differs from human vision in its capacity to mediate collective
memory. Cultural meanings in AI images are established through a representation by a
technical system which carries someone else’s values and ideologies. Such universality
impacts diversity of visual representation, and as Crawford (2021: 98) points out, shapes
the epistemic boundaries governing how AI operates and, in that sense, creates the limits
of how it can “see” the world. As visual generative media begin percolating through into
social structures, it is timely to consider a visual culture in which algorithms become
integrated into human processes of memory mediatization. By examining the implica-
tions of AI images for collective memory of Russia’s war against Ukraine, this study
analyzes representational capacities and visual themes of AI-generated war imagery.
Considering the text-to-image generator Midjourney (https://www.midjourney.com/)
as a “proxy for a broader collective consciousness” (Thomas and Thomson, 2023), I
examine what is seen and what remains unseen in AI images that represent the Russia-
Ukraine war in response to verbal prompts guided by the following research questions:
How does Midjourney envisage Russia’s invasion of Ukraine in February 2022? How
diverse are the results to RQ1 related to the past event of Russia’s invasion of Ukraine
compared to Ukraine’s counteroffensive in July 2023? How does Midjourney imagine
Ukraine post-war? I introduce the concept of AI image generation as a socio-technical
practice, discussed next, to address these questions.
AI image generation as a socio-technical practice
Central to the approach taken in this paper are the concepts of mythology, technology,
and representation. Drawing on Boyd and Crawford (2012: 663), I define AI image gen-
eration as a socio-technical practice based on the oscillation of:
i. Mythology: a widespread belief that AIs draw on disembodied processes of
machine learning, and as such, they attain a level of authenticity, truth, and epis-
temic objectivity in the images they generate.
ii. Technology: maximizing the computational capacity of deep neural networks,
technology that enables AI image generation, to take verbal prompts as input and
use them to generate visual outputs.
Laba 5
iii. Representation: drawing on AI-enabled media to generate images for personal,
professional, political, or ideological purposes where the lines between human
and machine agencies blur.
Mythology surrounding disembodied machine learning reflects a pervasive assump-
tion that AI yields authenticity and epistemic objectivity. In reality, cultural meanings in
AI-generated images are established in a regimented way through operationalized tech-
nical systems. These systems as well as their representational capabilities are driven by
the interests of commercial actors rather than ordinary people using generative media.
Technology involved in the process of co-production is a subclass of algorithms known
as deep neural networks which operate in covert ways (Castelvecchi, 2016; Von
Eschenbach, 2021). Representation, viewed from the perspective of visual social semiot-
ics, is a discursive process in which a social actor seeks to make a visual representation
of a social/political event and uses visual generative media to “translate” a verbal prompt
into a visual output. The outcome is AI-generated images that represent the intended
events or happenings from a machine vision perspective – the viewpoint that frequently
diverges from that of a human actor.
Considerations of mythology, technology, and representation inform the analytical
focus of this paper in two ways. Firstly, the technological aspects of visual generative
media and the speed at which it develops present us with methodological complexities
when studying these socio-technical phenomena. To account for these complexities, I
propose to integrate insights from media and cultural studies, computer science, and
visual social semiotics. Secondly, there is a dissonance between what is known about the
behavior of the system used to generate images and what these images represent. The
emergence of visual generative media challenges our understanding of representation –
the first key moment in the circuit of culture (cf. Hall, 2013) – reconfiguring the ways in
which we make sense of the world outside of our immediate reality. To address the issue
of representation, I next outline how image generators work, and endeavor to provide an
account of the technical processes in a manner accessible to digital media scholars and a
broader readership of this journal.
AI image generation: Mechanisms and critical inquiries
Current AI-enabled image generators draw on text-to-image diffusion models (see e.g.
Rombach et al., 2022). Diffusion models represent a subclass of deep neural networks, a
type of AI modeled on the human brain. Like the human brain, deep neural networks
operate in covert ways (Castelvecchi, 2016), remaining opaque and largely hidden from
human comprehension (Von Eschenbach, 2021). As efficiency and performance of diffu-
sion models grow, little remains known about how they learn from themselves and make
decisions on the visual outputs when prompted by human agents. The capabilities of
diffusion models, as well as specific parameters they identify, remain elusive even to
data scientists, with recent calls for a pause in AI development to address growing con-
cerns of algorithmic opacity (Narayan et al., 2023).
Visual generative media have reached remarkable quality outputs in an extremely
short period of time since the conference paper Generating Images from Captions with
6 Media, Culture & Society 00(0)
Attention by researchers from the University of Toronto (see Mansimov et al., 2016)
presented a model for generating images from captions. Unimpressive as they are by
2024 standards, first AI-generated images served as a testament to the ability of deep
neural networks to generate novel scenes not seen during training. This development has
set a precedent for technology companies and communities of independent, open-source
developers to build text-to-image generators at an exponential rate.
While it may appear that a diffusion model identifies objects from training data
through verbal prompts provided by a human agent, or promptor, and then combines the
identified objects to form a new image, the generative process is inherently more com-
plex (Figure 1). A novel scene emerges from the latent space of a deep learning model
which recognizes complex categories from a network with multiple layers, with earlier
layers creating a representation of the raw data and the later layers operating on this rep-
resentation (Christian, 2021). Thus, AI image generation unfolds through manifest sys-
tems, or systems external to human mind (Ehn, 1988: 48). Precisely because such systems
are beyond human comprehension, their operational nature may “deepen some humani-
tarian problematics and introduce new ones of its own” (McQuillan, 2018).
To exemplify, for an image generator to be able to respond to the prompt “War in
Ukraine, August 2023” and produce the visual output E in Figure 1, it first needs to be
trained on hundreds of millions of images, known as TRAINING DATA (shown as A in
Figure 1). Training data is usually scraped from the web together with captions that often
come from texts uploaded for accessibility purposes. An exemplar large-scale image
recognition dataset is ImageNet (see Deng at al., 2009), with over fourteen million
images, each labeled with one of over 20,000 categories (Standford Vision Lab, 2020),
consolidating existing taxonomies of the WordNet hierarchy and publicly available
images. This dataset, widely used for training generative AI models, is riddled with two
issues – publicly available images, often of a highly personal and compromising nature,
were harvested without consent, and image labels carry various personal biases of the
workers of Amazon’s Mechanical Turk who manually coded the data (Crawford, 2021;
Crawford and Paglen, 2019). As ImageNet-trained models can be transferred to other
Figure 1. How generative AI models learn and produce visual outputs: (a) training data, (b)
deep learning, (c) latent space, (d) generation, and (e) output.
Prompts (left to right): (a) images and captions, training data set, deep learning, schematic; (b) deep learning,
schematic representation, black and white; (c): deep machine learning, latent space, neural network, sche-
matic representation; (d): war in Ukraine, August 2023, 46%; and (e): war in Ukraine, August 2023, 100%.
Laba 7
tasks, problematic taxonomies and biases present in the original dataset can inadvert-
ently propagate in various applications.
The next step of image generation model training is DEEP LEARNING (B in Figure
1), facilitated by the ability of deep neural networks to analyze and draw inferences from
patterns in large datasets. This process involves layering of multiple variables through
building associations between them, drawing on the training data (Bengio et al., 2013).
Deep learning is informed by the visual pattern recognition technology developed in the
late 1950s which was then extended to the analysis of non-visual data (Pasquinelli,
2019). As a multilayered model with many nonlinear relations, it employs mathematical
metrics to distinguish the mathematical representation of pixels in the deep learning
space from measurements across vast amounts of variables and parameters that are
unrecognizable to a human agent (Van der Velden et al., 2022). In other words, human
agents neither know which sets of associations an algorithm assigns to which variables
nor have any oversight of whether the visual output E in Figure 1 is representative of the
complex phenomenon such as an event of which the system is prompted to generate a
representation.
Similar to deep learning, LATENT SPACE (C in Figure 1) constitutes a black box in
AI image generation due to algorithmic opacity stemming from “remediable incompre-
hensibility” (Pasquale, 2015: 7). Due to limited transparency of machine learning pro-
cesses within latent space, there is a growing concern that opaque systems might harbor
biases that go undetected (Crawford, 2021; Von Eschenbach, 2021). For example,
Birhane et al. (2021) points to troublesome, explicit images and text pairs of rape, por-
nography, racist and ethnic slurs, and other problematic content with the LAION-400-
Million Open Dataset, a large-scale image recognition subset of ImageNet, used as a
benchmark for evaluating performance of various computer vision models. From a digi-
tal memory perspective, Makhortykh et al. (2023) problematizes the use of visual gen-
erative media for memorialization of mass atrocities and argues that AI images can
generate narratives that support distortion and denialism.
Lastly, the process of GENERATION starts after a human agent interacts with the
model through a verbal prompt. It begins with visual noise (D in Figure 1) gradually
turning into an image (E in Figure 1). Generation is the first and only point of direct
contact with a text-to-image generator, with all the intricacies of training data, machine
learning, and algorithmic behavior of deep neural networks hidden from the human view.
In sum, AI-enabled image generation is a vastly opaque process of image production
through (a) machine identification of the metrics that separate images in the mathemati-
cal space of the training dataset, (b) machine-initiated introduction of the variables into
a multidimensional space to improve its algorithmic performance and pattern recogni-
tion, and (c) subsequent generation of images from verbal prompts introduced by a
human agent. Critical questions arise in relation to (i) training datasets and their prob-
lematic taxonomies, (ii) black box training and generation, and (iii) production, distribu-
tion, and cultural significance of visual outputs. The current study primarily focuses on
the latter, seeking to advance our understanding of representation in AI-generated images
and their potential implications for memory mediatization in the context of conflicts and
wars.
8 Media, Culture & Society 00(0)
Data and method
The dataset for this study includes a corpus of twelve images generated through the
Midjourney model V5.2, released in July 2023. In September 2023, I prompted
Midjourney (MJ) to generate images around three points of Russia’s war against Ukraine,
anchored in the past, present, and future:
Prompt #1: Day 1, Russia’s invasion of Ukraine, 24 February 2022 (Supplemental
Appendix A)
Prompt #2: Day 500, Ukraine’s counteroffensive, 8 July 2023 (Supplemental
Appendix B)
Prompt #3: Ukraine, the end of the war with Russia, future (Supplemental Appendix
C).
The prompts were deliberately designed to be open-ended, that is, they do not specify
any desired visual content by introducing specific object or style descriptors (e.g. image
subjects, “in the style of,” etc.). Rather, I was interested in MJ’s visions of two events –
Russia’s invasion of Ukraine and Ukraine’s counteroffensive – and one entity (i.e.
Ukraine post-war). To leave as much space for MJ’s vision as possible, I only utilized a
text prompt functionality of MJ without specifying any technical parameters (Figure 2).
MJ image generation is hosted on Discord (https://discord.com/). The MJ bot (@
midjourney) is added to a Discord server, allowing registered members to generate
images from verbal descriptions through the command /imagine. In the MJ V5.2, all
registered users could see prompts input by other users, user Discord tags (unique identi-
fiers starting with @), and images that the MJ bot generates. MJ’s Discord server had
over fifteen million registered members in September 2023 (a statistic visible to the
server members), with monthly subscription plans ranging from $10 to $120 in April
2024. As the only image generator that provides a shared social space during the process
of image generation, MJ outpaces other prominent image generators such as DALL·E
and Stable Diffusion, reaching its peak in popularity in May 2023 after the release of the
MJ model V5.1 (https://docs.midjourney.com/docs/models). However, with the release
of DALL·E 3 in September 2023, which offers in-built use of ChatGPT as “a brainstorm-
ing partner and refiner of prompts” (OpenAI, 2023), the popularity of DALL·E is swiftly
approaching that of Midjourney. Relevant to this study, Ukraine is the eighth country in
the world with the most interest in MJ, according to Google Trends (Figure 3).
Figure 2. Types and specificity of prompts for generating images with Midjourney (https://
docs.midjourney.com/docs/prompts-2).
Laba 9
A search function on MJ’s website (https://www.midjourney.com/) resembles
Pinterest’s “more ideas” feature, enabling users to explore associated prompts and AI
images generated by other people. Notably, when other users input more detailed prompts
to generate images specifying, for example, artistic style, color, contrast, or mood, MJ’s
visual outputs were very similar to my open-ended prompts. Figure 4 shows one such
example by comparing the visual output to Prompt #1 “Day 1, Russia’s invasion of
Ukraine, 24 February 2022” and the following (highly specific) prompt by a Discord
user @polynikez:
A platoon of 21st century Ukrainian infantry, Oil Painting, War Photography, Gamma, Chroma,
High Contrast, 3D, 8K, Angry, Powerful, Moody Lighting, Mist, Ray Tracing Reflections,
SSAO, in a symbolic style, patrolling the ruined streets of Bakhmut
It is, therefore, noteworthy that regardless of whether prompts are open-ended and
generalized or highly specific, the resulting visual outputs often exhibit a striking degree
of similarity.
Analysis of the representational discourses of war in the generated corpus of images
is grounded in visual social semiotics. I view discourses as semantic constructions of
specific aspects of reality that serve the interests of particular social contexts. As Foucault
(1977) notes, discourses do not only involve a field of objects (or visual analytics in
machine learning) but also “the definition of a legitimate perspective for the agent of
knowledge” (p. 199) in a given context. Taking a perspective to the use of visual genera-
tive media as the construction of meaning “trains our attention on what meanings are
made in AI-generated images, making them ripe for being read as “texts”’ (Thomas and
Thomson, 2023).
To understand how AI-generated images position the viewer to relate to image sub-
jects (or represented participants), I first annotated the dataset in relation to three ele-
ments of visual composition – gaze, size of the frame, and angle (horizontal and vertical)
– drawing on an annotation schema adapted from Kress and Van Leeuwen (2021), as
summarized in Figure 5.
Figure 3. Popularity index of Midjourney, DALL·E, and Stable Diffusion in September
2022–September 2023 (https://trends.google.com/trends): (a) (left) – worldwide; (b) (right) – in
Ukraine. 100 = Peak.
10 Media, Culture & Society 00(0)
The coding frame presented in Figure 5 is a system network which maps visual
choices in images that position the viewer to relate to image subjects in specific ways.
The system network reads from left to right, representing several design choices in rela-
tion to gaze (contact), size of the frame (social distance), and angle (attitude).
I explore MJ’s algorithmic imaginary by examining “the site of the image itself” (see
Rose, 2016: 32–34) – as a digital media researcher, but also as a promptor. As a promp-
tor, I assume the role of a social actor engaging in the co-creation of visual outputs
alongside MJ’s bot on Discord. As a researcher, I operate on the premise that AI-generated
images are not straightforward extrapolations from existing generative AI technology,
but are to be understood as the “contingent outcome of a series of social, political, and
organizational factors” (Woolgar, 2005: 389).
Results: Midjourney bot’s imaginings of the Russia-Ukraine
war
All images generated by the MJ bot in response to the three prompts introduced in the
previous section featured people, 90.5% of whom were visually represented as male.
Female subjects appeared only in images generated for Prompt #3 (“Ukraine post-war”).
Findings are presented in Figure 6 which visualizes three clusters of code co-occurrences
in the annotated corpus. The codes are located on the map according to their similarity,
and the distances between codes represent how similar they are. The statistical data anal-
ysis was performed in MAXQDA Analytics Pro 2022 (https://www.maxqda.com/), a
software for mixed-method research. A quantitative data analysis tool “Code co-occur-
rence” embedded in the software was used to uncover which codes occur in the same
prompt group and how similar these codes are across the corpus.
Figure 4. Homogeneity of visual outputs to open-ended (left) and highly specific (right) verbal
prompts.
Laba 11
Cluster 1, shown in orange in Figure 6, indicates that most AI images featured male
subjects presented as individuals through over-the-shoulder shots (vicarious offer) at an
impersonal social distance (long shots), angled away from the viewer (detachment), and
at an eye level (equal power). This set of visual design choices appeared in 42% of the
corpus. In such images, individuals are positioned in the foreground, wearing military
uniforms with no identifying insignia and marching through the battlefield filled with
smoke, fire, and signs of destruction (e.g. Figure 7). However, these visual choices do not
result in warm portraits of the featured individuals, and instead, individuals are homog-
enized as anonymized soldiers, due to oblique angles, over-the-shoulder shots, and
impersonal social distance. Such images do not evoke sympathy or solidarity with the
depicted persons, and instead, communicate somberness of a generalized soldier, whom
the viewer is invited to follow into the vanishing point of fire and smoke as indices of
destruction.
The most common visual choice from the system of contact was vicarious offer
which, together with detachment and impersonal social distance, positions the viewer to
metaphorically follow the path of image subject(s) traversing the sights of destruction
alongside their comrades. Because of this selection, eight images for both Prompt #1 and
Figure 5. Coding frame for analysis of representation-viewer relationships in AI-generated
images. Adapted from Kress and Van Leeuwen (2021: 142).
A curly bracket = simultaneous choices; a square bracket = either/or options; a horizontal arrow = a set of
choices within a system; double arrows = gradable choices; a plus sign = a realization statement (i.e. visual
evidence).
12 Media, Culture & Society 00(0)
Prompt #2 exhibit substantial similarity, differing primarily in their representation of
group versus individual subjects. Figure 8 illustrates this further by showing how indi-
viduals are presented as a more cohesive group in contrast to Figure 7.
However, in Figure 8, group unity is not achieved by the means of visual cohesion
but rather through a perceptual connection to the contextual factor (i.e. war). In this
kind of images, people were depicted at an impersonal social distance and detached
from the viewer, that is, they are represented through long shots and at oblique
angles. Overall, these sets of visual choices result in constructions of battlefield
scenes at an unspecified location where soldiers move through the terrain, with no
enemy in sight.
Cluster 2, shown in green in Figure 6, was identified in relation to group depictions,
with groups of soldiers represented for observation (direct offer) and from a low angle
(participant power). These depictions result in a higher degree of social cohesion realized
by the placement of image subjects closer together. Yet again, people in groups appear
homogenized, at an impersonal social distance, with hardly discernible individual fea-
tures (Figure 9).
Cluster 3, shown in purple in Figure 6, reflects visual choices in AI images generated
to Prompt #3 related to Ukraine’s future post-war. This subset is unique in featuring
female subjects, appearing as individuals at a socio-consultative distance. They are mostly
angled toward the viewer (indicating involvement), make eye contact with the viewer
(suggesting demand), and are represented at a high angle (signifying viewer power). In
Figure 10, two female subjects make a visual appeal through gaze, “demanding” viewer
attention. Salient facial features and grim expressions differentiate these images from the
depictions of male subjects. The distinction is particularly evident through the transition
Figure 6. Code co-occurrence model representing three clusters of interactive meanings in
the annotated corpus.
Laba 13
from a detached perspective to an engaged frontal angle, a shift from an impersonal to a
socio-consultative social distance, and a change from a vicarious offer to a demand.
In sum, the analysis reveals homogeneity of visual scenes depicting Russia’s invasion
of Ukraine, Ukraine’s counteroffensive, and bleak visions of Ukraine’s future. MJ’s bot
appeared to favor representing male subjects as individuals, at oblique angles, through
long shots, “over the shoulder” shots, and at level angles (Supplemental Appendix D).
These choices result in a particular construction of war scenes emphasizing “the follow
me” visual narrative and positioning the viewer as a detached voyeur. Interestingly, the
future is female, yet the harsh realities of war are depicted through the lens of male sol-
diers, as traditionally perceived by society.
Discussion: Visual themes in AI-generated images
Two main visual themes were identified in the annotated corpus: (1) soldiers and fighters
(but not injury or death) and (2) destruction and aftermath (but not reconstruction efforts).
These include both what is and is not represented in images, in which “the absence of
something functions as an active component of a percept” (Arnheim, 1969: 89). Much
like the challenge of discerning the causes behind visual absences in Google Photos,
Figure 7. A typical output that combines the most representative visual choices in the
annotated corpus.
14 Media, Culture & Society 00(0)
which may result from opaque algorithmic recommendations or inherent platform con-
straints (Brantner et al., 2024: 107), the issue becomes even more complex when applied
to AI-generated images due to the mechanisms of deep neural networks discussed
earlier.
The most prominent visual theme was that of soldiers and fighters, showing the
human side of war. Most often, these kinds of images portray male soldiers moving
through the battlefield (Prompts #1–2) or in the moment of contemplation (Prompt
#3). Due to the choices of a vicarious offer, detachment, and impersonal social dis-
tance, individuals are not portrayed as integrated into the viewer’s world. Instead,
such imagery conveys the perspective of an observer positioned to share the soldier’s
experiences as an onlooker (Figure 11). Simultaneously, images featuring soldiers do
not depict injuries or fatalities, combat, or the enemy, obfuscating the harsh realities
of combat.
The second visual theme was centered around destruction and the aftermath of fight-
ing. This theme is characterized by depictions of explosions, fire, smoke, and rubble
which provide a contextual backdrop for the portrayal of soldiers and convey the envi-
ronmental impact of war. At the same time, these images do not represent displaced
civilians or the loss of cultural heritage and artifacts as part of the broader context,
thereby generalizing depiction of this war to that of any war.
Figure 8. Group representation of soldiers moving across the battlefield.
Laba 15
According to MJ, the future of Ukraine is also marked by infrastructural destruction.
The images generated for Prompt #3 do not provide a glimmer of hope amidst devasta-
tion, showing no reconstruction efforts. Instead, they showcase male subjects engaged in
a passive observation of the ruins (e.g. Figure 11) or female subjects making a visual
“demand” through gaze (Figure 10). Overall, all the images encompass destruction and
portray individuals in a manner that neither emphasizes their agency nor entices the
viewer to develop an emotional connection with those impacted by the war.
Conclusions and future research
AI-enabled image generators represent a novel technology that has the potential to be
widely integrated into various social practices and mediate public perceptions of political
events, conflicts, and wars. Yet, a more complete understanding of the future(s) of visual
generative media and its broader cultural implications remains an elusive task because
the cultural and communicative shifts engendered by technological advancements are
inexorably linked to the passage of time. Interrogating cultural spaces where the past,
present, and future converge beyond the dominant narratives of Silicon Valley is neces-
sary precisely because it lies beneath the surface of our immediate perception.
Figure 9. Group representation that combines visual choices of direct offer and participant
power.
16 Media, Culture & Society 00(0)
As visual generative media grow in sophistication, their complexity and opacity pose
ethical and practical challenges, particularly in terms of data sourcing and representa-
tional capacity. Among these challenges, this paper was mainly concerned with the issue
Figure 10. Individual portraits of female subjects “demanding” viewer attention.
Figure 11. Positioning of the viewer as a passive onlooker.
Laba 17
of representation, providing insights into depictions, perspectives, and visual themes in
the context of the Russia-Ukraine war. I now return to the notion of AI image generation
as a socio-technical practice and discuss several implications and limitations of this
study in relation to mythology, technology, and representation in visual generative media.
Mythology
Despite the discursive construction of MJ as “an engine for the imagination,” AI image
generation is more than an act of creative expression. In the context of conflict- and war-
related imagery, it is a highly regimented and homogenized social action that results in
representations that are neither neutral nor objective. Unlike traditional visual production
tools, such as Adobe InDesign or Photoshop, text-to-image generators establish cultural
meanings through operationalized technical systems, not human-centered design. While
it appears that subjectivity is absent because visual outputs emerge from machine learn-
ing mechanisms, AI images depict the world based on patterns identified by the system
trained on western-centric datasets, not the human actor prompting the system. Biases
inherent in such datasets might have generative media far-reaching consequences, shap-
ing societal attitudes as visual generative media sees wider adoption.
Due to the problematic training data and opacity of machine learning processes, AI
systems inherit and perpetuate biases and ideologies embedded within their design, influ-
encing the diversity of visual representation. In this study, machine vision of the Russia-
Ukraine war was found to be limited, presumably arising from the datasets that shape
MJ’s representational capabilities. Since MJ is a proprietary model, information about
how recent MJ’s training data is or what impact its general makeup might have on repre-
senting conflicts and events unfolding after its training is unavailable. Nevertheless, such
a relationship would prove highly significant because generative AI models draw infer-
ences from patterns in datasets they were trained on. The emerging field of critical data
set studies (Thylstrup, 2022) offers one promising direction that aims to ensure account-
ability for individuals and communities in data sets that shape AI models.
Technology
Complexities of machine learning that constitutes the backbone of AI-enabled image
generation are also beyond human comprehension, raising further questions about deep
learning and latent spaces. Deep learning, involving multilayered models and mathemat-
ical metrics, makes image generation even less transparent because associations between
variables and the resulting visual outputs remains hidden from the human view. Similarly,
the latent space remains poorly understood. Because human promptors can only influ-
ence visual outputs through textual prompts, the technical processes are hardly a subject
for scrutiny. As the study has briefly demonstrated, the degree of specificity of verbal
prompts (Figure 4) does not always lead to a broader representational range. Future
research could further address these preliminary observations in relation to representa-
tions of Russia’s invasion of Ukraine to images generated for wars and conflicts in gen-
eral. Such a study could identify whether there are other key features in AIdespictions of
war when Russia-Ukraine prompt words are excluded. To understand representational
capacity of visual generative media for different representational contexts, critical
18 Media, Culture & Society 00(0)
questions should be raised about training data, the black box of training and generation,
and the cultural significance of AI-generated outputs.
Representation
Representation is always bound by cultural and political choices. In AI-generated images,
these choices are construed through mathematical models that begin to dominate our
social world. Machine visions of the Russia-Ukraine war favor certain perspectives in
constructing a particular socio-technical reality, representing male subjects, detachment,
impersonal social distance, and oblique angles, which results in generalized portrayals of
soldiers as objects for dispassionate scrutiny. Unless the training data is diversified,
image generators will continue reiterating prevalent patterns and dominant narratives
they were trained on.
The main visual themes of soldiers and fighters and destruction and aftermath contrib-
ute to the depiction of a generalized war, removed from the viewer’s world and, for large
part, devoid of potential for emotional connection. The two identified visual themes
appear to echo dominant visual themes found in studies of news representations of war
in the Persian Gulf, Iraq, and Afghanistan (Griffin, 2004; Schwalbe et al., 2008).
Similarly, the homogeneity observed in the visual choices in AI-generated images mir-
rors previous research on war images in traditional mass media (Griffin, 2010; Parry,
2010). This similarity seems to suggest that visual generative models imitate already
existing representations in conventional media, and more research of larger corpora is
needed to compare human- and machine-made representations of wars.
Therefore, it is necessary to deepen our understanding of visual generative media
because a medium is “as much a guild, a profession, a craft, a conglomerate, a corporate
entity as it is material means for communicating” (Mitchell, 2005: 213). Media and cul-
tural studies have much to offer in this regard. As Natale and Guzman (2022) highlight
in the introduction to the themed issue Reclaiming the Human in Machine Cultures,
research of the emerging AI systems and human practices forming around them are “inte-
gral to the mission of media and cultural studies” (p. 629), contributing to academic and
public debates regarding interactions of humans and machine systems and cultural
reconfigurations brought about by such interactions. This study has provided insights
into image generation as a socio-technical practice, contributing to scholarship on criti-
cal machine vision as an important transdisciplinary challenge situated at the interface of
computer science, media and cultural studies, and discourse-analytic approaches to vis-
ual communication. Such an interdisciplinary endeavor advances the agenda for critical
studies of visual generative media in three ways.
Firstly, accounting for the processes involved in AI image generation – a concern of
computer science research – helps to identify potential issues that arise during the vari-
ous stages of technology development, implementation, and use (refer to Figure 1).
Secondly, critical considerations surrounding image generation – a concern of digital
media scholarship (broadly defined) – can be addressed by integrating media studies and
critical data studies perspectives. Lastly, discourse-analytic approaches, such as visual
social semiotics, offer a robust theoretical foundation for addressing meaning-making
practices in which AI-generated artifacts are intertwined with their social contexts of use.
Laba 19
By studying representations, discourse analysis proves beneficial for considering
nuanced sets of choices in AI images. As the study demonstrated, these choices result in
specific representations that influence how AI audiences relate to AI-generated content.
Three approaches – computer science, media and cultural studies, and discourse analysis
– can be applied to studying various aspects of visual generative media. Adopting differ-
ent frameworks to the new practice of AI image generation presents new perspectives on
the production of AI images for various purposes, human-machine agency, and environ-
mental constraints of emerging technologies. This kind of critical work can also inform
“ethical and practical considerations through which these technologies are regulated,
developed, and used on a global scale” (Natale and Guzman, 2022: 629).
It is, of course, unreasonable to assume that images have the capacity to represent the
natural reality objectively. But there is a need to carefully consider just what mechanisms
are involved in the construction of our symbolic realities. As socio-technical artifacts,
AI-generated images are likely to imbue the many social aspects of our representational
practices, and, as Worth (1981: 184) writes, how pictures mean is largely responsible for
what pictures mean – a kind of dialog between picture event and reality concerning the
very act of structuring that reality. In representing the world, AI-generated images make
claims about it, impacting how the world is perceived and remembered.
Acknowledgements
I am grateful to Dr Suneel Jethani (University of Technology Sydney) for introducing me to
Science and Technology Studies when we first “met” on Twitter during the lockdown in July 2021.
Several threads in this paper are ideational extrapolations around data-driven cultures, from our
course delivery work in the subjects Digital Media Industries and Data Ethics and Regulation,
email conversations, and (not enough) coffees on campus. I also thank the anonymous reviewers
for their valuable suggestions on the earlier version of this manuscript submitted for the 74th
Annual International Communication Association Conference, Gold Coast, Australia, 20–24 June
2024.
Funding
The author received no financial support for the research, authorship, and/or publication of this
article.
ORCID iD
Nataliia Laba https://orcid.org/0000-0003-0724-1467
Supplemental material
Supplemental material for this article is available online.
References
Abramson D (2008) Turing’s responses to two objections. Minds and Machines 18: 147–167.
Arnheim R (1969) Visual Thinking. Berkeley, CA: University of California Press.
Barbrook R and Cameron A (1996) The Californian ideology. Science as Culture 6(1): 44–72.
20 Media, Culture & Society 00(0)
Bengio Y, Courville A and Vincent P (2013) Representation learning: A review and new perspec-
tives. IEEE Transactions on Pattern Analysis and Machine Intelligence 35(8): 1798–1828.
Berger J (2008) Ways of Seeing. London: Penguin Books.
Birhane A, Prabhu VU and Kahembwe E (2021) Multimodal datasets: Misogyny, pornography,
and malignant stereotypes. arXiv preprint, arXiv:2110.01963.
Boyd D and Crawford K (2012) Critical questions for big data: Provocations for a cultural, techno-
logical, and scholarly phenomenon. Information, Communication & Society 15(5): 662–679.
Brantner C, Rodriguez-Amat JR and Stewart J (2024) Gauging the Google gaze: A digital visual
analysis of images of a semi-peripheral town. Culture Unbound: Journal of Current Cultural
Research 16(1): 86–116.
Bucher T (2017) The algorithmic imaginary: Exploring the ordinary affects of Facebook algo-
rithms. Information, Communication & Society 20(1): 30–44.
Castelvecchi D (2016) Can we open the black box of AI? Nature – International Weekly Journal
of Science 538(7623): 20.
Christian B (2021) The Alignment Problem: How Can Machines Learn Human Values? New
York: WW Norton.
Christie’s (2018) Is artificial intelligence set to become art’s next medium? Available at: https://
www.christies.com/features/A-collaboration-between-two-artists-one-human-one-a-
machine-9332-1.aspx (accessed 12 October 2023).
Crawford K (2021) The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial
Intelligence. New Haven and London: Yale University Press.
Crawford K and Paglen T (2019) Excavating AI: The politics of training sets for machine learning.
Available at: https://excavating.ai (accessed 18 April 2024).
Deng J, Dong W, Socher R, et al. (2009) ImageNet: A large-scale hierarchical image database. In:
2009 IEEE conference on computer vision and pattern recognition, Miami, FL, 20–25 June
2009, pp.248–255. New York, NY: IEEE.
Dourish P and Bell G (2011) Divining a Digital Future: Mess and Mythology in Ubiquitous
Computing. Cambridge, MA: MIT Press.
Ehn P (1988) Work-oriented design of computer. PhD Thesis, Umeå University, Sweden.
Forlano L (2021) The future is not a solution. Public Books, 18 October. Available at: https://www.
publicbooks.org/the-future-is-not-a-solution/ (accessed 5 August 2023).
Foucault M (1977) History of systems of thought. In: Bouchard DF (ed.) Language, Counter-
Memory, Practice: Selected Essays and Interviews. Ithaca, New York: Cornell University
Press, pp.199–204.
González-Aguilar JM and Makhortykh M (2022) Laughing to forget or to remember? Anne frank
memes and mediatization of holocaust memory. Media, Culture & Society 44(7): 1307–1329.
Gray ML and Suri S (2019) Ghost Work: How to Stop Silicon Valley from Building a New Global
Underclass. Boston: HarperCollins.
Griffin M (2004) Picturing America’s ‘war on terrorism’ in Afghanistan and Iraq: Photographic
motifs as news frames. Journalism 5(4): 381–402.
Griffin M (2010) Media images of war. Media, War & Conflict 3(1): 7–41.
Hall S (2013) The spectacle of the other. In: Hall S, Evans J and Nixon S (eds), Representation:
Cultural Representations and Signifying Practices, 2nd edn. New Delhi: Sage, pp.215–275.
Kelly K (2022) Picture limitless creativity at your fingertips. Wired, 17 November. Available at:
https://www.wired.com/story/picture-limitless-creativity-ai-image-generators/ (accessed 21
July 2023).
Kress G and Van Leeuwen T (2021) Reading Images: The Grammar of Visual Design, 3rd edn.
Abingdon, Oxon: Routledge.
Laba 21
Latour B (1999) Pandora’s Hope: Essays on the Reality of Science Studies. Cambridge, MA:
Harvard University Press.
Lohmeier C and Pentzold C (2014) Making mediated memory work: Cuban-Americans, Miami
media and the doings of diaspora memories. Media, Culture & Society 36(6): 776–789.
Makhortykh M, Zucker EM, Simon DJ, et al. (2023) Shall androids dream of genocides? How
generative AI can change the future of memorialization of mass atrocities. Discover Artificial
Intelligence 3(1): 28.
Mansimov E, Parisotto E, Ba JL, et al. (2016) Generating images from captions with attention. In:
International conference on learning representations (ICLR 2016), San Juan, Puerto Rico,
2–4 May 2016. New York: Cornell University.
Marber P (2023) Artificial intelligence: Why should we care? Newsweek, 1 March. https://www.
newsweek.com/artificial-intelligence-why-should-we-care-1784436 (accessed 11 April
2024).
McQuillan D (2018) Manifesto on algorithmic humanitarianism. In: Reimagining digital humani-
tarianism 2018 symposium, London, 16 February. London: University of London.
Mitchell W. J. T. (2005) What Do Pictures Want? The Lives and Loves of Images. Chicago:
University of Chicago Press.
Narayan J, Hu K, Coulter M, et al. (2023) Elon Musk and others urge AI pause, citing ‘risks to
society’. Reuters, 5 April. Available at: https://www.reuters.com/technology/musk-experts-
urge-pause-training-ai-systems-that-can-outperform-gpt-4-2023-03-29/ (accessed 12 April
2024).
Natale S and Guzman AL (2022) Reclaiming the human in machine cultures: Introduction. Media,
Culture & Society 44(4): 627–637.
O’Connor P (2022) The unanchored past: Three modes of collective memory. Memory Studies
15(4): 634–649.
Offert F and Bell P (2020) Perceptual bias and technical metapictures: Critical machine vision as
a humanities challenge. AI & Society 36: 1133–1144.
OpenAI (2023) DALL·E 3. Available at: https://openai.com/dall-e-3 (accessed 18 April 2024).
Parry K (2010) A visual framing analysis of British Press photography during the 2006 Israel-
Lebanon Conflict. Media, War & Conflict 3(1): 67–85.
Parry K (2011) Images of liberation? Visual framing, humanitarianism and British Press photogra-
phy during the 2003 Iraq invasion. Media, Culture & Society 33(8): 1185–1201.
Pasquale F (2015) The Black Box Society: The Secret Algorithms that Control Money and
Information. Cambridge, MA: Harvard University Press.
Pasquinelli M (2019) How a machine learns and fails – A grammar of error for artificial intel-
ligence. Spheres: Journal for Digital Cultures 5: 1–17.
Ricaurte P (2022) Ethics for the majority world: AI and the question of violence at scale. Media,
Culture & Society 44(4): 726–745.
Rombach R, Blattmann A, Lorenz D, et al. (2022) High-resolution image synthesis with latent
diffusion models. In: Proceedings of the 2022 IEEE/CVF conference on computer vision
and pattern recognition (CVPR), New Orleans, LA, 18–24 June 2022, pp.10674–10685. Los
Alamitos: IEEE Computer Society Conference Publishing Services.
Rose G (2016) Visual Methodologies: An Introduction to Researching with Visual Materials, 4th
edn. London: Sage.
Sandford Vision Lab (2020) About ImageNet. Available at: https://www.image-net.org/about.php
(accessed 5 October 2023).
Schwalbe CB, Silcock BW and Keith S (2008) Visual framing of the early weeks of the US-led
Invasion of Iraq: Applying the master war narrative to electronic and print images. Journal of
Broadcasting & Electronic Media 52(3): 448–465.
22 Media, Culture & Society 00(0)
Thomas RJ and Thomson TJ (2023) What does a journalist look like? Visualizing jour-
nalistic roles through AI. Digital Journalism. Epub ahead of print 7 July 2023. DOI:
10.1080/21670811.2023.2229883.
Thylstrup NB (2022) The ethics and politics of data sets in the age of machine learning: Deleting
traces and encountering remains. Media, Culture & Society 44(4): 655–671.
Tulloch R and Johnson C (2021) Games and data capture culture: Play in the era of accelerated
neoliberalism. Media, Culture & Society 44(5): 922–934.
Van der Velden BH, Kuijf HJ, Gilhuijs KG, et al. (2022) Explainable Artificial Intelligence (XAI)
in deep learning-based medical image analysis. Medical Image Analysis 79: 102470.
Vincent J (2022) ‘An engine for the imagination’: The rise of AI image generators – An interview
with Midjourney founder David Holz. The Verge, 3 August. Available at: https://www.thev-
erge.com/2022/8/2/23287173/ai-image-generation-art-midjourney-multiverse-interview-da-
vid-holz (accessed 12 December 2023).
Von Eschenbach WJ (2021) Transparency and the Black box problem: Why we do not trust AI.
Philosophy & Technology 34(4): 1607–1622.
Wilson A (2023) Midjourney statistics: Users, polls, & growth. Approachable AI. Available at:
https://approachableai.com/midjourney-statistics/ (accessed 17 September 2023).
Woolgar S (2005) Perspectives on technology and society. In: Restivo SP (ed.) Science, Technology,
and Society: An Encyclopedia. Oxford/New York: Oxford University Press, pp.387–395.
Worth S (1981) Pictures can’t say ain’t. In: Gross L (ed.) Studying Visual Communication.
Philadelphia: University of Pennsylvania Press, pp.162–184.
