As AI image generation continues to grow, there’s chatter about the technology dominating the realm of digital representation. But can AI images ever truly represent us?

By Ashvaria Rai - Last Updated 11 April 2025

From Enheduanna to today's book marketers on TikTok, it’s no secret that the media landscape has drastically transformed throughout the ages. Digital platforms have especially accelerated a big paradigm shift towards a more “dynamic, open and participatory media environment” (Stevenson, 2), meaning content has become more accessible, instantaneous, and widely shared than perhaps ever before. At this change’s forefront, we find artificial intelligence (AI) technology.

(Embed: https://open.spotify.com/show/5C0th2EzmjtyutbdYBkU5R?si=7f1cd5b3e6ae471b)

AI is nothing new: As early as the now-ancient 1945 essay, As We May Think, AI was proposed as a system to further one’s knowledge and understanding of the world (Gunter, 5). But current feelings on AI are mixed as users explore both its potential and dangers. AI image generators, in particular, are a hot medium full of both opportunities and limitations. But which side outweighs the other? Let’s explore this through my AI selfie journey!

It's Not You, AI, But... It's Also Not Me?

The AI products of different prompts in different platforms.

Earlier this term, I experimented with AI imaging tools so you wouldn't have to. I wanted to include the following elements of my identity in the digital selfie:

Indo-Canadian
Queer 🌈
Love of Pop Culture (namely Harry Potter, Marvel as a close second)
Religion (Hinduism, specifically) 🛕
Writer 🖉
Physical features of tattoos, red and black hair, and a nose ring

To say the experience was frustrating is an understatement. Below is a recap of each app’s pros and cons and my final ranking (based on how strongly I resonated with the generated image):

We pretend that this is a table created in (and optimized for) Blogger, instead of a blurry photo insert.

Ultimately, none of the apps satisfied my want for an accurate digital representation - I simply had to concede and choose the least worst representation (DALL-E's version) as my consolation prize.

Digital Vs. Physical Representations

If selfies are an extension of personal identity (Liu et. al), then AI-generated selfies are a deeper exploration of our past, current, and hopeful future selves.
However, digital representations can also make us prone to identity performance or, as Goffman and Hogan addressed it, dramaturgy. We carefully curate our digital selves according to the online spaces we occupy. For example, I may create a more political identity on X vs. a more superficial and fun one on Instagram:

(Please pretend these are clear and embedded social media posts optimized for Blogger)

AI tools allow us to take this careful crafting one step further through personalized prompts and specialized digital choices. But, these tools won’t work for everyone in the same way. Interestingly, the supposed freedom of AI tools felt more limiting to me than traditional representations, and it’s obvious why.

Nature vs. Nurture: AI Biases = Media & Cultural Biases

Nature (Programming)

My AI selfie journey suggested how AI outputs are directly influenced by data biases in a Western-centric world (Hunter, Noble, Park), along with the cultural implications present in the resulting digital media. For example, womanhood is seen as existing in dichotomies (i.e. skinny/fat) rather than as a spectrum (Tiidenberg). Intersectionality, too, has a history of being erased, as overlapping factors of minority identities are often overlooked (Crenshaw, Fenner) in favor of creating one-dimensional caricatures (Ibid.). All media is thus shaped by culturally specific values, beliefs, practices, and politics (Stevenson, 3), and my stereotypical and whitewashed AI selfies reflect this.

“You can’t be Indian AND queer AND a writer! That’s absurd! Pick one struggle!!” – Gemini AI, probably (if it could talk)

Ogres are like Onions are like Intersectional Identities.

A bit ironically, AI has also been confined to dichotomous labels in media (Donna Haraway’s “A Manifesto for Cyborgs” nods somewhere in the background); it is often portrayed as a villain or hero in stories, rather than a tool used by people within that morality spectrum (Benson). In the current communications world, we may even see it only as a stealer of jobs, as the death of creativity, as the spreader of misinformation (Rubin), etc.

Nurture (Training)

But, reshaping our understanding of AI can help present it in a more accurate and relatable light that empowers online spaces and digital representations. One example of this is Seed, a WebToon working towards that more neutral depiction:

It seems super intense and anti-AI, but the trailer is misleading!

Along with better AI understanding, we can also still create more positive representations for minorities online. The first step is acknowledging AI's limitations and how political, societal, and economic biases play into AI content. Remember what Safiya Noble accomplished!?

(Embed: https://open.spotify.com/episode/3cP38Y1HiTeFFq8bBtmD20?si=eVJukeZxQLmxgNLCHltmJQ)

While AI is an inherently different medium, it can still be trained to better understand the complexities of humans. After all, Wu et al. were able to train AI models on abstractive book summarization! Yes, digital identities are more complicated, but studies have shown that AI can be trained through human feedback to diminish its programmed biases (Ananya, Manzoor). That’s why I implore you to try out these AI tools; by using them, we can train them to do better and contribute to a more inclusive and safe digital media landscape. Will training AI eliminate biases in all media? Of course not. But it’s a step in the right direction – and one we can all consciously take, especially those of us who find our digital identities compromised with current AI tools.

BONUS READ: Author's Notes on Transliteracy & More

"Transliteracy is the ability to read, write, and interact across a range of platforms, tools, and media from signing and orality through handwriting, print, TV, radio and film to digital social networks” - Sue Thomas

Transliteracy is a unifying concept practiced everywhere from India to France, in everywhere from Twitter posts to library websites (Thomas).

Admittedly, it was a bit difficult translating my academic thoughts to this blog. It involved a lot of deconstructing of what I think it means to be an academic (formal, polished writing) and allowing myself to be more open to experimentation, fun, and memes for a course assignment. The message certainly shifts when we go from an academic paper to the multimedia space of a blog, namely in tone and delivery, as the arguments become less detailed, more relatable, and more widely accessible. I feel more ownership of my findings as I can relay them in a more casual way that reflects more of my personality; I also enjoy making the findings more public-friendly through informal writing and multimedia rather than sticking to double-spaced size 12 font. McLuhan’s “medium is the message” (Bobbitt) has never resonated harder as the post creates a wondrous self-fueling cycle of knowledge and creativity for me:

McLuhan’s point of media being an extension of our “inner thoughts, ideas, and feelings—that is, an extension of inner consciousness” (Bobbitt) furthers my appreciation for this multimedia blog space, as it better encapsulates the more nuanced, layered, and creative side of my digital representation that the AI selfies did not. It also reflects on the beauty of transliteracy and being able to move between media forms to translate our messages for varying audiences. Transliteracy is not without its challenges, from its ethos complexities to the anxieties of creators (like myself) trying to enter transliterate spaces. Still, transliteracy ties us to further discussions on media, cyberspace, and technology, like Technobiophilia (Thomas).

(Embed: https://open.spotify.com/episode/2rGoLM8CPkjGCL5OQZJKKt?si=fa8e7omcQnuUQ6YEAC13nA)

Creating this blog post also helped me better understand how I fit into the role of AI, ethics, and culture - something I did not fully comprehend in my paper. By addressing the public, and encouraging them to act on AI bias, I got to reflect on my own opportunities (through work, research, and informal discussions) to combat AI and media bias in an everyday capacity. In this way, the blog medium can serve as a call-to-action, not only for the blog audience but for the creators (me) themselves too! The only downside I found to writing a blog post as opposed to the paper was sticking to a more concise length and falling within the word count (although, I'm not great with that in papers either!). However, the use of headings allowed me to break up my text and (hopefully) make it easier to engage with. Also, it seems impossible to embed audio/social media posts/tables in Blogger, which was a limitation to work around. Overall, I feel like this assignment made me feel more comfortable and excited for transliterate spaces, and develop an appreciation for the nuances and possibilities of training AI content generators! I'd be curious to try the AI selfie journey again in a year or so - hopefully to find more accurate results, as AI technology becomes "more than [they] were programmed to be" (The Wild Robot).

COMM 555: New Media Narratives

Search This Blog

If It Ain't White, It Probably Won't AI Right!