Crowdsourcing the Past

Citizen science can generate enormous volumes of transcription data — but volume alone doesn't make a good HTR model. This roundtable brings together practitioners working at very different scales, from tightly curated volunteer projects to large open platforms, to compare strategies for quality control, community engagement, and downstream model performance. The discussion addresses not just the technical questions — how to design workflows that serve both scholarly and computational ends — but also the human ones: how to co-design models with communities, recognise contributors' work, and use transcription projects as a vehicle for building genuine AI literacy beyond the archive.