Thread · 4 tweets · 28 Oct 2021

There might be too much to unpack in this for Twitter, but a few notes. First, antitrust doesn't mean splitting at the FB/IG/WA boundary — that's just reverting acquisitions. As @jhugman points out, various interop-based approaches are likely better (handwaving lots of details).

Crucially, different systems with different recommendation pathways is important so that we have editorial diversity. Reed's Law cuts both ways: 1bn people getting QA pushed to them and another 1bn getting some other madness is a lot less dangerous than 2bn getting QA.

Second, stopping surveillance doesn't mean going contextual. That's the maximalist view that the only alternative to invasive practices is to do nothing. You know from CI that that's not true! There is a whole set of parameters to apply to use but not abuse data.

Things like retention time, curbs on profiling, limits to AB testing, not reusing content provided for other purposes (eg. mining pics or 2FA numbers), keeping contexts separate, etc.