This bill is about enhancing the protections of lawful free-speech under the US constitution.

Generally speaking, we deem various kinds of speech that harms people as NOT protected under the first amendment, and that kind of speech would not be protected here.

Yelling “Fire!” in a crowded theater, libel and slander, speech calling for violence, and fraudulent advertising are some typical examples of speech not protected by the first amendment.

It would be tricky, but we could reasonably categorize engagement algorithms with certain properties as harmful to people and not subject to first amendment protections. This would be consumer protection, like laws against fraudulent advertising and other misleading claims.

For what it's worth, yelling fire in a crowded theater is First Amendment protected speech.

https://www.techdirt.com/2023/03/14/setting-1st-amendment-my...

I'm just stating facts.

https://www.popehat.com/p/the-first-amendment-isnt-absolute

> Generally speaking, we deem various kinds of speech that harms people as NOT protected under the first amendment, and that kind of speech would not be protected here.

That is not a general principle. Consider the following: it is now illegal to say the word “election” because it would harm the President.

The exceptions you're listing are mostly real but much more narrow than you're thinking. Only threats of imminent violence are excepted; it's completely protected by the First Amendment to say things like "it's getting to a point where we might have to be violent" or "I would commit violence if suchandsuch were to happen to me", even if other people reasonably find those statements to be stressful and alarming. Similarly, while defamation as such is illegal, it's not defamatory to say something like "John is a stupid fool who can't be trusted", even if I know many people think he's smart and trustworthy and even if his reputation is ruined by my statement.

There's simply no general principle that speech which harms people can be restricted.

> we could reasonably categorize engagement algorithms with certain properties as harmful to people and not subject to first amendment protections

This is a hope, however I have not seen any effort that wasn't scuppered, until recently.

The social media bans are a lurch in that direction.

I could separate them out into different strands:

1) Society saying that the engagement algorithms is not what we want to have in our lives or the lives of our children

2) Problematic technical implementations, or benign technical implementations that invade privacy and support government gaining more powers over speech.

Any regulation shaping algorithms is perilously close to shaping speech. Now if you say algorithms are speech or editorial decisions that platforms have their own freedom to choose, then you essentially strip them of their protections.

This would force a form of moderation that would have most people up in arms on HN.

I fully admit, I am being rough in my cuts; the point I am attempting to make is that any decision to decide what constitutes protected and unprotected speech is going to be government interference.

It may even be likely that the firms such as Meta or Tik Tok would end up as untenable under such rules.

The case which originated the "yelling fire in a crowded theater" analogy, Schenk v. US, was overturned more than half a century ago. And in that case, the proverbial fire-yellers were socialists distributing pamphlets.