Microsoft Research
Community Library Creator supports disability communities in defining good representation and curating images that reflect their perspectives. Developed through three months of collaboration with three disability organizations across the Global North and South, the project explores how community-led data practices can inform the evaluation and future adaptation of AI image-generation models.
During my Microsoft Research internship in 2025, I helped implement and iterate on multiple versions of the Community Library, analyzed workshop data through visualizations, and participated in workshops with the partner disability organizations.
- Published at the ACM CHI Conference on Human Factors in Computing Systems, 2026. Paper Link
- Photo pinboards and magazine activities help communities articulate what meaningful representation looks like.
- Community libraries organize selected images and annotations.
- AI image evaluation brings community perspectives into assessments of generated representations.
Community Library Creator workflow. Figure 2, Thieme et al., CHI 2026.
Defining representation through a community-curated photo pinboard. Figure 6, Thieme et al., CHI 2026.
Evaluating generated images against community-defined representation. Figure 5, Thieme et al., CHI 2026.
Engaging Communities Meaningfully in Defining Disability Representation for AI Image Generation
- Human-Computer Interaction
- Accessibility
- Generative AI
- Disability Representation
- Community-Led Data Curation
Anja Thieme, Rita Faia Marques, Martin Grayson, Sidhika Balachandar, Cameron Tyler Cassidy, Madiha Zahrah Choksi, Camilla Longden, Reeda Shimaz Huda, Nicholas Ileve Kalovwe, Christina Mallon, Courtney Mansperger, Daniela Massiceti, Bhaskar Mitra, Ruth Mueni Nzioka, Ioana Tanase, Yuzhe You, Cecily Morrison.