Clean

[edsu]

When pointed at a JSONL file Hydrator could rehydrate or clean the dataset, filtering out any tweets that have been deleted. In addition the Hydrator could allow user to enter an optional list of users to exclude from the dataset, if someone has requested their data be removed, and they don't want to force the user to delete or protect their tweets.

(Thanks to @bergisjules for the use case and the name).

[edsu]

Here's some more context for this issue.

On a recent DocNow call we discussed whether it would be possible for users to register that they wanted to be removed from a tweet id collection. The Hydrator could then consult that list when hydrating the ids, and filter them out.

However we wouldn't want people to have to publicly register that they want to be removed since it could bring unwanted attention to their choice. But there are cases when we might want the registration to be public (e.g. elected officials).

It would be relatively easy to allow a Hydrator user to enter in a list of users to exclude from hydration. But are there any creative approaches for allowing the user to know what Twitter users might want to be removed?

[edsu]

I was thinking that perhaps one obvious approach would be not to publish the User IDs of the users who wish to be removed, but to publish hashes of them instead.

When the Hydrator hydrates the data it could then hash the User ID and see if matches one that wants to be removed and act accordingly. The problem then is that other people could easily write a tool that walks through the dataset and highights that the user wanted to be removed.

So hashes by themselves aren't really enough. It feels like there is some missing social piece that can help ensure contextual integrity.