AI Wont Show My Disability My Story

AI: A Powerful Tool and a Personal Disappointment

I genuinely love Artificial Intelligence. It's a remarkable tool with the staggering potential to transform how we communicate, create, and connect with one another. From enhancing accessibility for individuals with disabilities and enabling businesses to operate more efficiently, to unlocking novel forms of creativity, AI is actively redefining possibilities across every sector. However, a recent experiment involving AI generating images of myself led to an unexpected and stark realization.

AI could produce a sophisticated, professional image of me, or even depict me enjoying a sunny vacation. Yet, there was a significant, glaring omission: the AI consistently gave me two arms, instead of one.

Embracing Disability as Identity

I was born without my left arm. While growing up, I encountered challenges many might not comprehend, I also cultivated a strength and resilience that have become fundamental to my character. I proudly represented Australia as a Paralympic swimmer and now dedicate myself to advocating for the rights of people with disabilities. My personal story isn't about overcoming my disability; it's about embracing it as an essential component of my identity. It profoundly shapes my worldview and fuels my dedication to fostering a world where every form of diversity is acknowledged and cherished.

You can read more about similar perspectives in the essay, How I Came to Love My Bionic Hand.

When I ask an AI to generate an image, I want it to reflect the entirety of who I am—including my disability. My disability is not a flaw to be concealed or corrected. It is an integral part of my identity, my strength, and my narrative. Until technology truly grasps this, there is significant work ahead.

Understanding AI Bias and Its Impact

AI bias arises when artificial intelligence systems yield prejudiced or inequitable outcomes. This can stem from biased training data, flawed algorithms, or a deficiency in diverse representation during their design phase. When technology neglects to encompass the full spectrum of human diversity, it risks reinforcing stereotypes and exacerbating societal divisions.

A prominent illustration of AI bias was observed in facial recognition technology, which was found to misidentify individuals with darker skin tones at alarmingly higher rates. In 2018, a study conducted by MIT Media Lab highlighted that commercial AI systems exhibited error rates as high as 34% for darker-skinned women, in stark contrast to a mere 1% for lighter-skinned men. This significant disparity underscored how a lack of diverse data and representation in AI development can precipitate detrimental and biased results.

More Than Just an Image: Erasing Disabled Voices

My experience is a single illustration of how disabled voices can be marginalized when inclusivity is not integrated into the design process from the outset. This isn't solely about my personal experience; it pertains to the wider issue of disability being omitted from the narrative—both literally and figuratively.

AI possesses the capability to dismantle barriers and augment accessibility, but only if diversity and inclusion are embedded at its very core. This necessitates the active involvement of disabled individuals—not as an afterthought, but as pivotal contributors to the innovation process.

A Troubling Reflection and a Desire for Specificity

What I found most disheartening was the realization of the immense potential we had to alter this narrative from its very inception. Instead, AI is currently acting as a mirror to our society, reflecting its inherent biases.

Ironically, my initial apprehension regarding AI was that it might become "too intelligent" or "too specific." Now, however, I find that specificity is precisely what I desire. I do not want generalized experiences or interactions; I want AI to be customized to me, accurately reflecting who I am.

I decided to prompt the AI to generate an image with explicit instructions to include my disability. Even with these direct prompts, it appeared to be an overly challenging task for the AI to simply reflect my true self in its generated images. Consequently, I asked the AI why it found it so difficult to create an accurate image of me.

The article notes that several figures mentioned in the original piece illustrate this point:

• Screenshots from a conversation Smith had with ChatGPT, with AI-generated images, all courtesy of Jessica Smith.
• Additional AI-generated images, also courtesy of Jessica Smith, are described as part of her experience.

These examples further underscored the AI's difficulty in accurately representing her, even when explicitly asked.

A Pivotal Moment: Shaping AI for True Representation

We stand at a juncture with an opportunity to reshape this narrative. As AI continues to be developed and refined, now is the critical time to construct it through the lens of diversity and disability. This is a crucial moment where we can influence the course of deeply embedded social constructs. It is an extraordinary and exciting prospect—but it requires immediate action.

Imagine if AI-generated images routinely depicted people using wheelchairs, prostheses, or with any other visible sign of disability, without any special request. Imagine if AI offered this representation to everyone by default.

Beyond Reflection: AI as a Shaper of Inclusive Norms

How remarkable would it be if we could train AI to actively challenge societal perceptions of disability—not merely upon request, but as an inherent part of its design? The bias often emerges from the assumption that everyone desires to "improve" their image, and disability is frequently overlooked as an option because it is persistently viewed as "less than."

Disability is not a problem requiring a solution, but a valuable perspective to be embraced. By integrating lived experiences into the design process, AI can evolve beyond merely reflecting societal norms to actively reshaping them. This can lead to the creation of a world where every individual is seen, heard, and represented authentically.

The Future We Can Build: AI for All

AI should not just mirror our current society; it should be instrumental in forging a more inclusive one. The outcome of such an endeavor is not merely superior technology, but a more equitable society where innovation truly serves everyone.