AI-Shaped Self: When the Identity Created Is Not Fully Me - Dan Yuan

 

AI-Shaped Self: When the Identity Created Is Not Fully Me

In this era of rapid AI development, creating an AI-generated selfie is a very interesting experience. By describing the version of myself in my mind, AI generates what it “understands” as me based on these descriptions. This process made me rethink something I had always taken for granted: whether my self-image is truly completely under my control.

At the beginning, I treated AI simply as a tool that executes instructions. However, during the creative process, I gradually realized that no matter how detailed and specific my descriptions were, the final generated image could not fully reflect what I had in mind. The person created is not entirely “me”, it is more like a product co-created by me and the AI system.

Creative Process

Figure 1. Powered by ChatGPT

When creating the AI selfie, I chose to use ChatGPT as the image generation tool. One reason is that I am using the Plus version, so I also had certain expectations for its output quality. During the process, I did not focus on making the image look exactly like me, because for me, expressing emotion and atmosphere was more important.

Figure 2. My cat, symbolizing emotional comfort in my project

In my prompts, I described feelings such as loneliness, homesickness, and the experience of celebrating Chinese New Year alone as an international student. At the same time, I included many symbolic elements. For example, my cat represents emotional comfort, snow symbolizes the coldness and sense of distance of the city, and lanterns along with Chinese New Year elements represent family and cultural memory. These elements helped me express not physical accuracy, but my inner emotional state.

Figure 3. AI-generated self-representation that merges present reality

However, the final generated image did not completely match my expectations. Although the overall atmosphere was consistent with what I intended, there were differences in the details. The image appears more idealized than my real emotional experience, and even somewhat exaggerated. This made me realize that AI is not simply executing instructions, but interpreting and reconstructing them in its own way.

AI and My Identity

After repeatedly modifying my prompts and trying to make the image closer to what I imagined, I gradually realized that my self-representation is not fully under my control, but is shaped together with the AI system. The “me” generated by AI feels like a simplified and slightly idealized version.

Figure 4. AI-generated selfies

Some complex or contradictory emotions seem to gradually disappear in the final image, replaced by more obvious and even slightly exaggerated emotional expressions. As a result, the image becomes easier to understand, but it also loses its subtlety. This suggests that AI is not only representing identity, but also shaping the identity it “understands.” Relying on this way of expressing the self over time may gradually influence how we understand ourselves.

Rethinking AI Narratives and New Media

After creating these images, I found that they resonate with the concept of “missing narratives” proposed by Chubb et al. (2022). Many current narratives about AI tend to focus on the future or extreme scenarios, while overlooking the everyday experiences of ordinary people. My project attempts to highlight more relatable emotions such as loneliness, homesickness, and cultural identity.

However, even these personal expressions are still influenced by AI systems. Even when we try to express real emotions, they are reorganized according to the aesthetic logic of AI. This means that these narratives are not entirely authentic, but are partially constructed by technology.

Manovich (2001) also argues that digital media does not simply reflect reality, but constructs it. In my images, different times and spaces exist simultaneously: the snowy city represents my current life abroad, while the warm family gathering symbolizes memory and imagination. These elements cannot coexist in reality, but are combined into a single image through AI, creating a layered identity composed of reality, memory, and emotion.

From Academic Essay to Blog

An important part of creating this blog was translating my academic analysis into a form more suitable for online reading. In my original paper, I used more formal language and focused more on theoretical analysis. In contrast, in this blog, I intentionally used simpler expressions and included more personal experience, using “I” to reduce the distance between myself and the reader.

Compared to the academic version, this blog focuses more on storytelling rather than complex theoretical explanation. This shift also made me realize that changes in medium directly influence how meaning is communicated. As McLuhan suggests, “the medium is the message.” Writing in a blog format not only makes the content easier to understand, but also changes how readers receive and experience it. Compared to academic writing, this form feels more like telling a personal story rather than simply conducting analysis. Overall, this project made me realize that AI-generated selfies are not simple reflections of the self, but complex results shaped by both human intention and technological systems. While AI opens up new possibilities for self-expression, it also reshapes and simplifies personal experience to some extent. Therefore, it is important to critically reflect on how AI influences identity in digital media.

Reference

Chubb, J., Reed, D., & Cowling, P. (2022). Expert views about missing AI narratives: Is there an

AI story crisis? AI & Society, 39, 1107-1126. https://doi.org/10.1007/s00146-022-01548-2

Manovich, L. (2001). The language of new media. MIT Press.