Thread by @daphnehk

Some initial thoughts on this "JAMA" bill repealing 230 immunities for algorithmically targeted content.
energycommerce.house.gov/sites/democrat…

It's odd to me that the bill targets *personalized* ranking but not general engagement-based ranking. If the problem is that ranking is personalized, you would think the legislative fix lies in privacy rights rather than changing a content law like CDA 230. 1/

As usual, the kind of prohibited ranking sweeps very broadly, and would seem to discourage platforms from making beneficial ranking changes in response to e.g. coordinated inauthentic behavior. (Maybe the idea is to address that by covering only personalized ranking?) 2/

The scienter element is wacky. Liability comes not from knowing/reckless disregard about illegal content. It comes from knowing/etc. about the basic, omnipresent fact that the service is ranking third party content. Drafting error or deliberate? I assume the latter. 3/

Rather than tying this to existing, known, illegal things, this covers recommendations that "materially contributed to a physical or severe emotional injury to any person." That... seems to me to run into Item #1 on my list of six const. hurdles. 4/ cyberlaw.stanford.edu/blog/2021/01/s…

That part is especially weird paired with the scienter part, where it seems like the platform does not need to have known or had reckless disregard about the harm itself, or the content itself. 5/

I really like that it has exemptions for small businesses and for infrastructure providers. That should be part of platform regulation drafting 101. 6/

I don't know that this small business definition is the right one. I mean, I very seriously don't know. I think a real effort by economists and Internet experts to quantify this before we make any laws about it would be incredibly useful. 7/

The infrastructure carve-out has a nice list of examples. But like another bill (PACT, maybe?) it weirdly only applies if the DNS/caching/CDN/etc. is acting as a back-end service provider for a *different* interactive computer service. 8/

I assume the idea is that someone, somewhere in the technical stack should hold liability for illegal content. But that disregards other big issues, including the bluntness and inaccuracy of removals from things like AWS (taking down a whole website for one bad comment, e.g.). 9/

Anyhow... Bottom line is that I am not too surprised to see this. I'm sure you aren't either. 10/

It's also a big flashing sign: Anyone who thinks (1) amplification causes harms, but also (2) any regulation needs a lot of nuance: Your work is done on (1).
If you believe in (2) – the nuance part – it is time to start talking about that. Or we'll get laws like this one. 11/11

Oops. I failed to plug my article about regulating amplification. It goes into way more detail, especially on First Amendment issues. knightcolumbia.org/content/amplif…

Good point from @BerinSzoka. The "with respect to information" phrasing causes two problems. First, as he points out, stripping (c)(1) immunity 100% opens the door to more "must-carry" claims from -- and more platform concessions toward -- the far right.

https://twitter.com/BerinSzoka/status/1448691903013302276?s=20

Second, I think it means that the inevitable false or erroneous allegations about newsfeed content lead to removal from the entire service. Bc the platform loses immunity for that piece of "information" anywhere it sits.

And let's not even get into what "information" means and whether this opens up claims that platforms must filter for "identical" or "equivalent" information.

If we do go there, though, man does the EU have some lessons for us. This is a whole narsty kettle of fish.

If you want to know more about must-carry claims, my take on the Con Law issue is here, in 2nd half: lawfareblog.com/who-do-you-sue…

And @ericgoldman has a piece, I think with @jess_miers, showing how often these are resolved on grounds of 230(c)(1). NOT (c)(2).

Pay attention, Dems. This (c)(1) / (c)(2) stuff matters for the substantive issues you care about.

@ericgoldman @jess_miers If you want to know the speech, privacy, and possibly equal protection problems with a law that de facto requires platforms to build filters...
The short version for a US audience is item #4 here: cyberlaw.stanford.edu/blog/2021/01/s…

The long version, in my best EU legalese, is here. academic.oup.com/grurint/articl…