A new paper published on arXiv introduces a taxonomy for understanding and mitigating task-level biases in historical image retrieval systems. The research highlights how AI models, when applied to digitized cultural heritage collections, can inadvertently amplify existing biases present in historical imagery and metadata. The paper examines the sources of these biases, such as differences between training data and historical images, image quality issues, and incomplete descriptions, and reviews current methods aimed at addressing these challenges. AI
IMPACT This research could lead to more equitable and accurate AI-powered search tools for historical archives and cultural heritage collections.
RANK_REASON The item is an academic paper detailing a new taxonomy for AI bias in image retrieval. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- GLAM institutions
- Gotit.pub
- Hugging Face
- Ratan Jagan Sebastian
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →