Recreating me: An exploration of identity through the production of AI selfies
Objective
In this assignment, I attempted to replicate different versions of myself through the creation of AI selfies. These selfies were meant to capture identities I felt I had portrayed digitally through platforms like Linkedin and Meta, which I challenged by creating two selfies that provided deeper insights into my personality and perspective.
This objective comes from the idea that people are multifaceted, and intentionally create specific restrictive or exaggerated representations of themselves.
Hogan (2010) takes Evring Goffman’s overview of how individuals have a front stage self that they portray to various audiences, and how we act to produce a certain image of ourselves, and considers how that applies to our participation in the digital world, ultimately framing that that behaviour extends to the way we represent ourselves on social media. In doing so, he analyzes how we project particular personas through social media by considering our audience when deciding which parts of our identity we choose to reveal or hide (Hogan, 2010).
My goal was to compare the selfies that represented these various identities with one that I attempted to make represent my understanding of my own identity, to better demonstrate the way we play to different audiences while exploring the possibilities and limitations of AI and the selfie in embodying those features.
Methodology
The basic task was to instruct AI to create selfies by feeding it the characteristics I felt I displayed on various social media I had shared to various platforms. I used Adobe Firefly for the first 3 images, and Gemini AI (Google) for the fourth.
Prompting AI required thoughtful and intentional descriptions of what I wanted - instead of saying “I am a reader”, I had to consider how that trait would be physically manifested, for example, by asking AI to have me hold a book. This took multiple attempts and fine-tuning, which often required more exact language, for example in my first attempt to make my photo representing myself on LinkedIn, I mentioned I wanted the Alberta Legislature in the background, but on a second try I had to specify that the building should be in the near background, as the first rendering had the building too far into the background to be easily recognized.
The first 3 prompts began with a consistent physical description, including brown wavy hair, brown eyes, and tan skin. For the selfies that were based on social media platforms, I had to go to my profile and activity on those platforms, and think about what traits of mine could be gained by looking at them.
Image 1: LinkedIn
For Image 1, which represented myself on LinkedIn, my prompt was comprised of the limited information on my LinkedIn such as my job (working as a policy advisor for the Alberta government, hence the Legislature), the fact that I’m a reader (represented by the book in my hand), that I like to represent myself as a professional (wearing a suit) and the idea that I enjoy my job (represent by my smiling).
Image 2: Meta
For Image 2, which represented myself on Meta, I looked at my Facebook and Instagram and found recurring themes such as travel, particularly to the mountains and lakes (I asked for mountains and water in the background). I often share pictures with friends for drinks or dinner, so I have a drink in my hand in the photo. I also tried to give the impression that I was fun and had an open and social personality (hence the large smile and inviting eyes).
Image 3: Real me
Image 3 represented a more private version of myself as I attempted to time larger emotional, social, and political issues that dominate my life. While I did try to add elements that represented my physical tendencies that are unknown to most, such as someone who often sits cross legged at home, the fact that I have lots of plants in my living room. Much of the work involved translating larger more complex themes such as spirituality (hence the Buddha statue), my need to stay informed politically or be constantly entertained by media (hence headphones), and a desire to use my talents (such as writing, hence the notebook).
After receiving feedback on my first round of selfies such as encouragement to refine my prompts for greater precision in order to better integrate more complex identity themes, I made a fourth selfie attempt.
Image 4: Real Me (Take II)
I created a fourth selfie with the intention of better representing precise elements that illustrated my lived experience and perspective. This involved switching from Adobe to Gemini AI (Google) as after experimenting with different prompts, I found Gemini to be better equipped to create complex images, such as by integrating more physical elements into each image.
This new software allowed me to create a more realistic rendering of my physical traits, better implementing more detailed elements of my physical description, including hair length, eye size, and eyebrow shape.
Larger emotional themes included:
At this point in time, the political state of the world, especially the relationship between Canada and the United States, takes up a lot of my time and emotional energy, which is why I asked AI to depict me looking nervously at a TV with the US flag on it, as if watching the news.
This political awareness and anxiety comes from a position of knowledge about and passion for government relations, which I represented through the degree hanging on the wall (though AI was not able to capture that it was a political science degree) as well as the various books scattered.
I embody my national identity through the use of the Canadian flag. This gives further insight into my political perspective. The flag’s use also aligns with a recent growth in my own Canadian pride, as many Canadians have expressed a rise in national identification in the wake of trade war with the US (Saba, 2025).
I use yoga, plant care, and the practice of Buddhism, in order to help me deal with the negative feelings associated with the political climate, which is why there are multiple yoga mats, plants, and buddha statues throughout the background.
None of these elements are things I typically share online, and yet these are the ideas that are most present in my day to day life.
Analysis
The use of AI for this assignment was beneficial in its convenience - plugging in and refining a prompt was much quicker than staging any of the above photos. However, what the images were intended to express was still subject to a number of barriers and complications, including limitations to the performance of AI, and limitations to the use of the selfie.
Limitations of AI
AI applications are not finished products, but developing and learning systems fraught with valid concerns.
Image Quality and Theme Integration
AI software can only perform as well as the user can guide it. Over the process of attempting to create suitable images, a lot of fine-tuning and re-wording was required. In addition, Adobe Express struggled with integrating too many elements at once, making it difficult to include too many details, resulting in more simple photos. Further, Adobe struggled with creating high-quality images. For example in Image 3, my criss-crossed legs have disappearing feet.
Algorithm, Bias, and Culture
AI is not immune to bias. Notions and images represented through AI are bound to the ideological inclinations of a system conceptualized and coded by biased individuals (Laba, 2024). As a result, AI has the capacity to carry and reinforce institutional norms and commercial interests (Laba, 2024).
Both AI softwares I used show a clear bias towards conventionally attractive people. While how attractive someone is is subjective, AI consistently produced images that displayed me as a relatively thin woman, with commonly sought-after features like full lips and large eyes.
Even using the phrase “I am of average attractiveness” while experimenting with my prompt made no difference to the physical details the AI put forth.
Not all bias is so easily identifiable as conventional attraction, many other elements of the photos produced could be a result of unknown algorithmic bias. If Ai is incapable of accurately representing reality due to bias, this could negatively impact the way we understand and consider the world at large, especially as the use of AI for rendering images grows (Laba, 2024).
Ethical and Cultural Concerns
All choices have the potential to be ethical choices, even the ones we don’t typically evaluate on a moral level - this remains true for AI.
As mentioned above, creating a good image through AI required multiple attempts - and with those attempts, a lot of brain power - which is concerning considering how the energy required to facilitate the use of AI remains an environmental issue (Ren and Wierman, 2024). Further, as users we rarely get a comprehensive overview of how AI models are trained (Laba, 2024), which makes it harder to truly know the ethical implications of the use of each software. This includes concerns regarding copyrights and property law, as exemplified in an ongoing lawsuit, targeting Meta for its alleged use of pirated books in order to train AI models (Milmo, 2025).
Limitations of Selfies
I ultimately discovered that selfies have the ability to communicate ideas and larger themes. Unfortunately, this communication was limited by the amount of information that could fit into one photo, and the traits of mine that could be illustrated in a tangible way. Further, choosing what to represent also meant prioritizing information that I felt would be relevant and impactful to disclose. This inevitably meant deciding what elements were not worth adding into the prompt. Ironically, this means a selfie can not possibly portray ‘the full picture’ of who someone is.
Social Media and Audience
Bullingham and Vasconcelos (2013) argue that the physical separation from others afforded to us in online spaces encourage individuals to represent themselves differently, while recognizing that even those representations are rooted in conventional norms.
The first two images I shared were bound to prompts based on what information was available on my social media.The limitations of this approach were most evident in my LinkedIn image, which had few intentional details because my LinkedIn does not actually provide a lot of information about me. My opportunistic use of LinkedIn, while intentionally revelatory of only a small sliver of my identity, is still tied to conventional norms as LinkedIn is often used to promote (rather than represent) oneself to a professional audience (Dijck, 2013). This is because LinkedIn users anticipate an audience of potential employers, clients, business partners, and professional connections and thus want to demonstrate an advantageous version of themselves (Hogan, 2010).
However, there is still a limitation to the photos where I deliberately attempt to better lay myself open to the viewer. Even though the class blog is fairly insular, I have no guarantees that any data I publicly share here can be scrubbed from the internet. As Hogan (2010) notes, online artifacts created in one context can move across time and location and thus be interpreted in a totally different context, which can impact how it is received. By creating and sharing a selfie like the ones I have created online, posters, including myself, leave themselves vulnerable to the unknown impacts of their permanency. Liu et al. (2018) argue that selfies are an opportunity to interact with societal norms. Under their perspective, by posting something with information I would not typically post on social media, especially in the context of its possible permanency, I explore the boundaries of what is or is not acceptable to share online in a non-anonymous format.
Outsider Interpretation
Without this accompanying text, the selfies I create could be understood completely differently. Not everyone has the background to make the linkages between the items I’ve selected and the messages I attempt to portray through them. Someone could see the image of the Buddha in my final photo and just think I really like little statues, rather than recognize them as a symbol of my faith.
Liu et al. (2018) recognize that selfies can have the ability to create insider and outside groups. By choosing elements in order to portray nuanced narratives, I potentially cast off those who do not have the context to make that connection as external to the reception of my intended message. However, by posting the selfies online I also have the ability to open myself to further connection with those who understand what I am attempting to communicate, especially if they choose to further participate (Liu et al, 2018) such as through leaving a comment.
Conclusion
Often, concerns related to AI are impacted by dramatic dystopian narratives that place the consequences of AI as far-off future imaginings, but this focus takes away from the many real challenges AI users face in the current day (Benson, 2022). As AI becomes more commonplace, it is crucial to understand its limitations and risks, such as bias, lack of quality production, environmental impacts and ethical concerns, and work to address these areas. Algorithmic bias is a particularly concerning aspect, as it has the ability to negatively impact culture and how we understand the world around us by reinforcing institutional norms and corporate agendas.
Additionally, selfies and the use of social media more generally are considerate of their audiences, and thus not always fully transparent. In addition, selfies are not always understood as they were meant to be shared, which can distort meaning or exclude individuals from the conversation the image attempts to initiate. As a society, we should seek to better understand each other by not relying entirely on digital representations, such as those found on social media, and rather grow discourse by adopting a variety of meaningful ways to connect with one another.
That is not to say that AI or selfies are inherently bad. Selfie creation provided me the opportunity to broadcast information and interact with norms. Thus, we should not demonize the use of AI or selfies, but rather ground their use in an understanding of their limitations in order to interact with them better.
Sources
Beson, David. (2022, June 3). AI in Fiction and the Future of War. The Strategy Bridge. https://thestrategybridge.org/the-bridge/2022/6/3/ai-in-fiction-and-the-future-of-war
Bullingham, L., & Vasconcelos, A. C. (2013). The Presentation of Self in the Online world’: Goffman and the Study of Online Identities. Journal of Information Science, 39(1), 101–112. SageJournals. https://doi.org/10.1177/0165551512470051
Dijck, J. van. (2013). “You have one identity”: performing the self on Facebook and LinkedIn. Media, Culture & Society, 35(2), 199–215. SageJournals. https://doi.org/10.1177/0163443712468605
Gunter, D. (2023). AI, the new frontier – opportunities and challenges - Research Information. Research Information; ResearchInformation.info. https://www.researchinformation.info/analysis-opinion/ai-new-frontier-opportunities-and-challenges/
Hogan, B. (2010). The Presentation of Self in the Age of Social Media: Distinguishing Performances and Exhibitions Online. Bulletin of Science, Technology & Society, 30(6), 377–386. SAGE Publications. https://doi.org/10.1177/0270467610385893
Laba, N. (2024). Engine for the imagination? Visual generative media and the issue of representation. Media Culture & Society, 46(8). https://doi.org/10.1177/01634437241259950
Liu, F., Ford, D., Parnin, C., & Dabbish, L. (2018). Selfies as Social Movements: Influences on Participation and Perceived Impact on Stereotypes. Proceedings of the ACM on Human-Computer Interaction, 1, 1–21. ResearchGate. https://doi.org/10.1145/3134707
Milmo, D. (2025, January 10). Zuckerberg approved Meta’s use of “pirated” books to train AI models, authors claim. The Guardian; The Guardian. https://www.theguardian.com/technology/2025/jan/10/mark-zuckerberg-meta-books-ai-models-sarah-silverman
Ren, S., & Wierman, A. (2024, July 15). The Uneven Distribution of AI’s Environmental Impacts. Harvard Business Review; Harvard Business Publishing. https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts
Saba, M. (2025, February 6). Canadian pride is on the rise in wake of Trump’s tariff threat, polls suggest. CBC. https://www.cbc.ca/news/politics/canadian-pride-is-on-the-rise-trump-tariff-threat-1.7451987
Comments
Post a Comment