Theory of the Selfie (Part 3) - Sarah Mate

 

A video compilation of my digitally created selfies.

In a time where our lives are dominated by digital connectivity, the selfie has emerged as a new form of popular media. As AI technology continues to become mainstream, it  emerges as a powerful tool with the potential of shaping digital identities.

AI art representing the act of taking a selfie in the style of a renaissance painting.

There are a myriad of ways one could create a digital selfie using AI. One is to use prompts to create a completely AI created image of yourself. Another is to use a current photo and apply various AI powered filters. These filters are now commonly used on social media, with users sometimes not able to recognize themselves. If the selfie has been filtered to the point it no longer looks like the subject, this is considered a form of catfishing. (Lavrence and Camber, 2021, p.3) While the research demonstrates users asking others if their photos look like them, it neglects to examine if people do know their filtered photos no longer look like them and are seeking reassurance for their altered digital identity. 

 

An example of a heavily filtered image. Source: BoredPanda

 

The Creation Process

Because I had previously used the free trials for many of the systems that enhance current images, I decided to use Microsoft Image Creator to create my Selfie (this service is available with a paid Microsoft account). This AI system generates images based on user inputted text prompts. While these tools can be exciting to use, they are not without fault, and have many ethical implications – specifically with a bias toward thin, white, cisgendered people and away from minority groups and marginalized communities. (Sharma and Grayson, 2021) 

The Microsoft Image Creator Interface

Analysis of Representation

I began creating my AI selfie by using basic prompts that describe me – 43 years old, woman, blonde, brown eyes, wears glasses. I noticed the images the system generated were consistently of a thin white woman. Academic readings affirm this bias with AI generated imagery. (Hunter, 2022) When I added I was overweight, the image produced was of a person much larger than me. The women in the digitally produced images all seemed either much younger or much older. I was not able to get the system to generate an image that accurately displayed my age or my body size.

AI generated art to display how polarizing the results can be for women. Notice the distinct lack of cultural diversity.

Incorporation of Feedback In the transliteration process, I incorporated both instructor and peer feedback. I elaborated on the concept of AI based photo filters, clarifying their role in perpetuating “catfishing” by “enhancing” images to the point of being unrecognizable. I included more scholarly resources to enhance the depth of analysis and added context about the inherent biases present in AI systems and their harmful effects on marginalized communities. I refined my critical analysis of sources, integrating them within the narrative to offer a more comprehensive examination.  Ethical and Cultural Considerations I struggle to use current AI art systems knowing the datasets used to power them are comprised of art created by artists who never provided consent for their art to be used for this purpose. Culturally, I witnessed first-hand the erasure of both women of colour (all images produced were of white women even though I didn’t specify race or ethnicity), and larger bodied people. On the flip side, marginalized communities such as the transgender community, have discovered the use of AI-enhanced selfies as beacons of empowerment. Through digital avatars, individuals can navigate and affirm their identities with newfound authenticity. (Roberts and Prisma, 2022). In examining the impact of AI generated and filtered selfies, it’s important to consider the research in this area is quite recent, and it’s difficult to fully understand the magnitude of how harmful these systems will be on marginalized communities over time. AI Generated Image Transliteracy Transforming an academic critical analysis paper into a blog post is a challenge. First, academic writing is very different than writing for the web. Ideally, web text is succinct and easy to read. Second, blog posts are typically only 500-700 words. We began this process with a 1200 word academic paper, incorporated feedback from both a peer and our instructor which added more information to the 1200 words, and then edited it down to the blog post you are reading now. Conclusion This process has been an interesting digital journey that allowed me to learn the value of using the correct prompts with AI tools, receiving and incorporating feedback, and transforming academic writing into a web-based blog post. The tactical aspects of this assignment had to happen alongside the research and theory aspects to ensure it was completed on time. While I am grateful to have the opportunity to flip back and forth between theory and tactical work, I wonder if this assignment allowed me to fully absorb the many aspects included as part of it.  References: Chayka, K. (2023, February 10). Is A.I. Art stealing from artists?. The New Yorker. https://www.newyorker.com/culture/infinite-scroll/is-ai-art-stealing-from-artists  Tiidenberg, K. (2018) Visibly ageing femininities: women’s visual discourses of being over-40 and over-50 on Instagram, Feminist Media Studies, 18:1, 61-76, DOI: 10.1080/14680777.2018.1409988 Laccetti, J. (2024) Module 4 Lecture Notes Lavrence, C., & Cambre, M. (2020). “Do I look like my selfie?”: Filters and the Digital-Forensic Gaze. Social Media + Society, 6(4), 205630512095518. https://doi.org/10.1177/2056305120955182   Hunter, T. (2023, April 14). AI selfies — and their critics — are taking the internet by storm. Washington Post. https://www.washingtonpost.com/technology/2022/12/08/lensa-ai-portraits/ Roberts, M., & Prisma, L. A. &. (2022, December 9). The euphoric highs & problematic lows of AI avatar art. Refinery29. https://www.refinery29.com/en-au/ai-avatar-art Sharma, S., Graydon, M. S., & NASA. (2021). Social Bias in AI and its Implications. In NASA STI Program Report Series. https://ntrs.nasa.gov/api/citations/20210010446/downloads/NASA-TM-20210010446.pdf