Everyone seems to be overlooking something here. Decent weeding software doesn’t just look at a library’s own catalog and usage stats- it can also compare a library’s holdings against a larger database (OCLC’s Worldcat, in the case of academic libraries) to determine whether a library’s copy of something is the only existent copy, whether there are other copies within the state that can be borrowed via interlibrary lending, etc. Obviously, there are limitations to the data, but algorithms CAN be a tool to help determine rarity.
