If I may offer yet another different reasoning, I think that data consumers are expected to have a minimum of data literacy.
In that sense, the “human effort” would be to clean up data that has quality problems, e.g., removing non-data headers from a spreadsheet, converting dates that are in non-standard format, fixing incorrectly marked character encoding. That is, the usual data cleaning jobs. Having to scrape websites or PDFs would also qualify as “human effort” for the effort in setting up these scrapers. Improper documentation or lack thereof (i.e. an API manual, a data dictionary explaining the meaning of fileds in a spreadsheet) is also a factor against making data usable and requiring effort to figure things out.
In my opinion, I think neither spreadsheets nor APIs should be penalized just for being so. Instead, this criteria should reflect the amount of hurdles faced by data literate users in actually making use of the data.