Filling the Gaps: AI, Identity, and the Power of Defaults (Assignment 4)

Filling the Gaps: AI, Identity, and the Power of Defaults 

Journey of Creation 

    When I first started this project, my goal was simple: to create a digital version of myself using AI. However, the process quickly became more complex, making me question how identity is translated into data and interpreted by algorithms. To create my digital self-representation, I used Craiyon and ChatGPT’s image generator with the same prompt, describing my appearance, clothing, and my dog, to compare how each system handled the input. 

    One detail became central to my analysis. When I mentioned that I had a dog but did not specify the breed, the AI consistently generated a Golden Retriever or a similar “friendly-looking” dog. After I refined the prompt and added the breed, the image changed (Figure 1). This showed that the system cannot leave gaps. It fills them using familiar and widely recognizable patterns. This made me realize that AI image generation is not just about what I input, but also about how the system fills in missing information. 

 

Figure 1, main selfie generated by ChatGPT

Figure 2, generated by Craiyon

Figure 3, generated by Craiyon

Figure 4, generated by ChatGPT

Self-Representation and Identity 

    The generated images reflected my identity but felt simplified, shaped by both my description and the system’s assumptions. At first, I thought this was due to uncertainty. Later, I understood that the system selects statistically common patterns when details are missing. The repeated dog example made this clear. A similar tendency appears more subtly in other aspects of representation, where undefined identity markers often align with dominant visual norms, as I observed in my own experiences, which tended toward a Caucasian appearance, even though I did not mention ethnicity. As McKinlay (2010, p. 235) explains, identity is produced through the “forced reiteration of norms.” In this sense, AI-generated images do not simply reflect identity; they participate in reproducing it through repeated patterns. 

Ethical and Cultural Dimensions 

    These patterns raise important ethical questions. AI-generated selfies may seem playful, but are shaped by datasets reflecting existing norms. As Sujon et al. (2025, p. 5) note, “selfies also occupy a complex position at the intersection between visual culture and datafication processes.” This means that AI selfies are not just images, but part of a system that structures how identity becomes visible. The issue is not just resemblance, but how the system defines what is normal. The dog and ethnicity examples show how quickly the system defaults to familiar representations when information is incomplete. 

Another aspect I considered was the choice of visual style. I avoided hyper-realistic images at first because of concerns about how realistic AI visuals can be used, especially in contexts like deepfakes. Instead, I chose a webtoon style. This created distance while also reflecting my identity, as I work as a translator of Korean webtoons. 

Conclusion 

    This journey showed that AI selfies are not simple reflections of identity. They construct recognizable versions of identity using familiar patterns. The repeated appearance of default elements highlights how these systems prioritize recognizability over specificity. As Chubb, Reed, and Cowling (2024, p. 1107) argue, “stories are an important indicator of our vision of the future.” Even small outputs like AI selfies contribute to shaping how identity and technology are understood. In this sense, the issue is not only the system itself, but also the data it is trained on and who has the power to produce and shape that data. These patterns reflect not just technical processes, but inequalities in representation and access.

Transliteracy Reflection 

    Writing this as a blog post changed how I presented my ideas. Compared to my academic paper, I focused more on clarity and readability. This reflects McLuhan’s idea that “the medium is the message,” where “the personal and social consequences of any medium... result from the new scale that is introduced” (McLuhan, 1964, p. 1). The format itself shapes how meaning is communicated. 

At the same time, this process helped me understand transliteracy. Adapting my argument into a blog was not just about simplifying it, but about restructuring it for a different medium. This showed me that ideas need to be reshaped rather than simply transferred across formats.

References

Chubb, J., Reed, D., & Cowling, P. (2024). Expert views about missing AI narratives: is there an AI story crisis?. AI & society, 39(3), 1107-1126.

McKinlay, A. (2010). Performativity and the politics of identity: Putting Butler to work. Critical Perspectives on Accounting, 21(3), 232-242. https://doi.org/10.1016/j.cpa.2008.01.011 

McLuhan, M. (1964). Understanding media: The extensions of man. McGraw-Hill. 

Sujon, Z., Iqani, M., & Jonathan, S. (2025). The Invisible Lives of Selfies. 

By Iffet Secil Kinsan