>Several times we ran A/B tests with many millions of people to try exactly this. Every time all the usage metrics tanked. Not just virality and doomscroll metrics, but how many likes, messages, comments, re-shares, and app-opens. We never even measured ad-related things on that team.

Well, yeah, but this has an implicit "engagement === good" assumption. Exactly the same thing that incentivizes unhealthy McDonald's food: they make more money when they sell food that still leaves you hungry. So, yeah, people probably did want this, and when they got it they started using Facebook in a healthy manner (no point opening it at every available moment to just scroll through 'new' trash), which tanked your metrics. If you're actually worrying about your users you should also consider that them using your product more might not actually be what they want or need.

Ironically enough, I think the same mistake (or rather, it's more of a mistake because there's not quite such a naked financial incentive to make this worse for the affected users) has happened with the youtube analytics dashboard: multiple youtubers have said that it's actively addicting and really bad for their mental health, but any change that feeds that probably looks really good in their metrics because, hey, creators are using it more, that must mean it's good, right?

Trust me, I came in there full of motivation for "do what is good for the actual humans", and most of the rank of file were the same. FB's employees are not evil or exploitative, though I won't say its unfair to describe the leadership in such terms.

Many times in product design meetings I would interject with "but this hurts people!" etc.

We hated that our personal careers were directly tied to increasing the junk-food factor. It didn't feel good at all. But the choice, as crafted by HR and senior directors was clear: Junk food this thing, or lose your jobs.