A new study published on arXiv explores how blind and low-vision (BLV) scientists interact with AI tools like ChatGPT and Gemini to understand multimodal scientific papers. Researchers interviewed five BLV and five sighted scientists to identify current practices, accessibility workarounds, and preferences for AI-generated responses to visual content. The study found that vague descriptions and incorrect AI outputs can lead scientists to abandon these tools, and it offers a dataset of 115 queries and responses to aid future research in developing more accessible AI-powered scientific QA systems. AI
IMPACT This research highlights the need for more accessible AI tools in scientific research, potentially influencing future development of multimodal QA systems.
RANK_REASON The cluster contains an academic paper detailing research on AI tools for scientific document analysis. [lever_c_demoted from research: ic=1 ai=1.0]
- AI-powered scientific QA systems
- Arnavi Chheda-Kothary
- arXiv
- blind, low-vision scientists
- ChatGPT
- Gemini
- multimodal scientific papers
- sighted scientists
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
