Our colleague Thomas Padilla, now at the Center for Research Libraries, is doing a survey as part of a new project called Ground Truths which is looking at the very important issue of using cultural heritage data in machine learning contexts (there are real problems with well-labeled training datasets, for example). If you follow this link, there’s more background on the project before you start the survey form. See
I think this should develop some very interesting data for the community, and as it’s compiled I’ll share pointers here. Also, hopefully, we can get an update on this work as it moves forward during one of our upcoming CNI meetings.