However, I think the court is attempting to draw a line between expressive editorial acts and non-expressive, instrumental harms. This distinction can be seen through an analogy:

Imagine a physical newspaper that included only third-party content but also chose to add an addictive substance to its physical pages. When a reader touches and turns the pages, it triggers a compulsive, dopamine-addiction response. In this scenario, should the act of adding the addictive substance—which has nothing to do with the expressive value of the content—be considered an act of a "publisher" that should be shielded by a law like Section 230?

My intuition is no and I believe this is what the court is trying to get at. The core of the complaint against Meta is that the design of the service (e.g., infinite scroll, autoplay) is engineered to be addictive, irrespective of the content itself. This design is an instrumental act of manipulation, separate from the act of "speech/speaking" that Section 230 was designed to protect as harm here is not derived from the substance of the third-party content item, but from the mechanism used to deliver it, making it a harm that is not truly related to – or caused by – the content that was published.

I’m interested to hear your thoughts and whether you think we should prefer protecting this type of deliberate act because of the potential impact on third party speech/content over incentivizing platforms to take responsibility their own goals to maximize time and attention spent on their platform.

I'm not sure that link came through and can't seem to edit my comment, so here's the full URL — the case is Chosen Figure LLC v. Kevin Frazier Prods., No. 22-06518, 2023 U.S. Dist. LEXIS 125669 (C.D. Cal. July 19, 2023): https://business.cch.com/ipld/ChosenFigureKevinFrazierProds20230719.pdf

Some online journals seem to target the content of an edition to the subscriber. The publisher is simply publishing multiple user-specific journals under the name.

A social medium platform differs from an online journal because the social medium platform holds out transport of a message under standard terms for compensation. In contrast, Scientific American only transports its content to a subscriber.

(I) By common carriage doctine (especially in MA and in CA), the social medium platform provides common carriage and publishing service. When the social medium platform is a common carrier of messages (a modern telegraph), it is not liable for the transported message. O’Brien v. W. Union Tel. Co., 113 F.2d 539 (1st Cir. 1940).

(II) A social medium platform usually does not create its own content and is not liable for other people's content, but moderation or curation is expressive action. The social medium platform seems potentially liable for the harmful effects of its expressive actions.

When the social medium platform tries to invoke the CDA for situation (II), it is dishonestly trying to hide its own expressive action to escape liability.

A social medium platform sometimes escapes liability via the CDA because the social medium platform mixes two logically distinct services. The courts could force clarity by requiring a social medium platform to fulfill its common carriage obligations in a message conduit layer. Then the social medium platform would have the option — if it wanted to be liable for its expressive action — of creating a separate moderated or curated message stream with full First Amendment protection.
https://uploads.disquscdn.com/images/5fb6cf5acc3aaa8c27a7e5ab520daae62203b2e3230a6a920bdbe43b522279b1.jpg