Thread by @cfiesler

I'm alarmed by this exact issue, and here's a related one I've thought a lot about:
All datasets that curate "public" data (e.g., photographs or social media posts) create secondary archives of content that otherwise the original content creators would have control over. [Thread]

https://twitter.com/random_walker/status/1318965082294988806

There are many reasons why you might want to delete content that you originally shared publicly, and why that content still being used by others (even scientists) might be harmful. One example that comes to mind is someone who has been through gender transition.

I also wrote about a speculative example in this design fiction about research ethics--in which there is a curated dataset of "last words" of deceased life-loggers, then used by trolls to harass their surviving loved ones. cmci.colorado.edu/~cafi5706/grou…

I think that respecting the right of a content creator to control visibility of their content when possible is important. AO3 designs for this; see the "History vs Control" subsection of this paper: cmci.colorado.edu/~cafi5706/CHI2…

The obvious counterpoint to this is, essentially, "but the data is public," and once you put something online you no longer deserve control of it. But the idea that nothing else matters but "publicness" (including how personal that content might be) is absurd.

This kind of thinking leads to absurdity like scraping incredibly personal content from a dating site and then releasing it as a curated public dataset without any anonymization. wired.com/2016/05/okcupi… (as described by @michaelzimmer)

This comes down to tensions b/W open science & research ethics/privacy. If I had to give recs:
(1) some datasets are too sensitive to be amplified even if they curate "public" content;
(2) sharing *by request* is an option, so you can vet research goals.
i.e.: CONTEXT MATTERS.