Painting with Cameras and Drawing with Text: AI Use in Accessible Creativity
Cynthia L. Bennett, Renee Shelby, Negar Rostamzadeh, Shaun K. Kane · 2024 · Proceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '24) · doi:10.1145/3663548.3675644
Summary
This paper explores how disabled creatives intentionally use and choose not to use generative AI (GAI) tools in their artistic practices. Through interviews with 10 creatives with disabilities — whose work spans audio engineering, leathercraft, visual art, photography, music composition, writing, and digital design, and whose disabilities include blindness, low vision, motor impairments, chronic pain, ADHD, and invisible disabilities — the researchers examine the intersection of accessibility and creative expression. The study is grounded in crip technoscience, a framework from disability studies that positions disabled people as expert knowledge-makers who generate innovative technical solutions from their lived experience of navigating inaccessible systems. Rather than framing GAI as either universally beneficial or threatening to disabled creatives, the paper reveals a nuanced landscape where participants strategically adopt, adapt, and refuse different AI tools depending on how those tools interact with their access needs, creative identity, and artistic values. The paper identifies three cross-cutting themes: access hacks (the creative workarounds disabled people develop to navigate inaccessible tools and environments), the entanglement of creative practice and access work (how the labour of creating access is inseparable from the creative process itself), and deliberate positioning of GAI within workflows (participants' thoughtful decisions about where AI helps and where it threatens their artistic agency).
Key findings
Participants described a rich ecosystem of access hacks developed long before GAI existed — a blind photographer who uses camera sounds and tactile feedback to compose shots, a leatherworker who adapted tools for one-handed use, a musician who developed keyboard macros to compensate for limited hand dexterity. These hacks represent genuine creative expertise, not deficits to be fixed. When participants adopted GAI, they used it in targeted ways: image generation to create visual reference materials they could describe to collaborators, text-to-image tools to rapidly prototype concepts that would otherwise require inaccessible visual software, AI transcription to caption their own audio work, and large language models to help structure and edit written content. Critically, participants were highly selective about where GAI entered their workflow — they typically used it for access-related tasks (overcoming inaccessible tool interfaces) rather than replacing their core creative decision-making. Several participants explicitly refused to use GAI for the parts of their practice they considered most artistically meaningful, even when AI could technically do the task faster, because the creative struggle and embodied process were integral to their art. The paper surfaces important tensions around authorship and attribution — when a blind artist uses text-to-image generation, is the resulting work "theirs"? Participants felt strongly that their creative vision, prompting skill, and curatorial judgement constituted genuine authorship. The study also found concerns about GAI training data excluding disabled perspectives, producing outputs that reflect ableist norms (e.g., image generators that default to non-disabled bodies), and the risk that GAI could be used to justify not hiring disabled creatives by claiming the AI can do the work instead.
Relevance
This paper is essential reading for anyone working at the intersection of AI and accessibility. It challenges two common narratives: that GAI is primarily a tool for fixing disability (the technosolutionist view) and that GAI threatens to replace human creativity. Instead, it shows that disabled creatives are sophisticated, intentional users who integrate AI as one tool among many in practices already shaped by decades of creative adaptation. For accessibility practitioners, the concept of access hacks is directly relevant — these represent design knowledge that often goes unrecognised. The paper's framework of "accessible creativity with responsible AI" proposes four qualities: GAI should support creative agency (not replace it), respect embodied knowledge, acknowledge the labour of access work, and be designed with disabled creatives as co-designers rather than afterthought users. Limitations include the small sample size and the rapidly evolving nature of GAI tools, but the conceptual contributions — particularly the entanglement of access work and creative practice — are durable insights that will remain relevant as specific tools change.
Tags: generative AI · creativity · disability · responsible AI · access hacks · crip technoscience · blind and low vision · motor impairment