Sarah Constantin wrote:

Specifically, I think that LW declined from its peak by losing its top bloggers to new projects. Eliezer went to do AI research full-time at MIRI, Anna started running CFAR, various others started to work on those two organizations or others (I went to work at MetaMed). There was a sudden exodus of talent, which reduced posting frequency, and took the wind out of the sails.

One trend I dislike is that highly competent people invariably stop hanging out with the less-high-status, less-accomplished, often younger, members of their group. VIPs have a strong temptation to retreat to a "VIP island" -- which leaves everyone else short of role models and stars, and ultimately kills communities. (I'm genuinely not accusing anybody of nefarious behavior, I'm just noting a normal human pattern.) Like -- obviously it's not fair to reward competence with extra burdens, I'm not that much of a collectivist. But I think that potentially human group dynamics won't work without something like "community-spiritedness" -- there are benefits to having a community of hundreds or thousands, for instance, that you cannot accrue if you only give your time and attention to your ten best friends.

While I agree that the trend described in the second paragraph happens (and I also dislike the effects), I have another model that I think more tightly explains why the first paragraph happened. I also think that it's important to build systems with the constraint in mind that they work for the individuals inside those systems. A system that relies on trapped, guilted, or oppressed participants is a system at risk for collapse.

So in order to create great public spaces for rationalists, we don't just need to have good models of community development. Those can tell us what we need from people, but might not include how to make people fill those slots in a sustainable way.

To explain why lots of top bloggers left at once, let me present a model of adult development, drawn from George Vaillant’s modification of Erik Erikson’s model, as discussed in Triumphs of Experience. It focuses on 6 different ‘developmental tasks,’ rather than ‘stages’ or ‘levels.’ Each has success and failure conditions associated with it; a particular component of life goes either well or poorly. They’re also not explicitly hierarchical; one could achieve the “third” task before achieving the “second” task, for example, but one still notices trends in the ages at which the tasks have been completed.

Triumphs of Experience is the popsci treatment of the Harvard Grant Study of development; they took a bunch of Harvard freshmen and sophomores, subjected them to a bunch of psychological tests and interviews, and then watched them grow up over the course of ~70 years. This sort of longitudinal study gives them a very different perspective from cross-sectional studies, because they have much better pictures of what people looked like before and after.

I'll briefly list the developmental tasks, followed by quotes from Triumphs of Expertise that characterize succeeding at them. The bolded ones seem most relevant:

• Identity vs. Role Diffusion: “Live independently of family of origin, and to be self-supporting.”
• Intimacy vs. Isolation: “capacity to live with another person in an emotionally attached, interdependent, and committed relationship for ten years or more.”
• Career Consolidation vs. Role Diffusion: “Commitment, compensation, contentment, and competence.”
• Generativity vs. Stagnation: “assumption of sustained responsibility for the growth and well-being of others still young enough to need care but old enough to make their own decisions.”
• Guardianship vs. Hoarding: The previous level covered one-on-one relationships; this involves more broad, future-focused sorts of endeavors. Instead of mentoring one person, one is caretaker of a library for many.
• Integrity vs. Despair: Whether one is graceful in the face of death or not. [1]

As mentioned before, they're ordered but the ordering isn't strict, and so you can imagine someone working on any developmental task, or multiple at once. But it seems likely that people will focus most of their attention on their earliest ongoing task.

It seems to me like several of the top bloggers were focusing on blogging because something was blocking their attempts to focus on career consolidation, and so they focused on building up a community instead. When the community was ready enough, they switched to their career--but as people were all in the same community, this happened mostly at once.

I think that “community-spiritedness” in the sense that Sarah is pointing at, in the sense of wanting to take care of the raising of new generations or collection and dissemination of ideas, comes most naturally to people working on generativity and guardianship. People work on that most easily if they’re either done with consolidating their career or their career is community support (in one form or another). If not, it seems like there’s a risk that opportunities to pursue earlier needs will seem more immediate and be easily able to distract them.

(In retrospect, I fell prey to this; I first publicly embarked on the project of LW revitalization a year ago, and then after about a month started a search for a new job, which took up the time I had been spending on LW. If doing it again, I think the main thing I would have tried to do differently is budgeting a minimally sufficient amount of time for LW while doing the job search with the rest of my time, as opposed to spending time how it felt most useful. This might have kept the momentum going and allowed me to spend something like 80% of my effort on the search and 20% on LW, rather than unintentionally spending 100% and 0%.)

1. I interpret the last task through the lens of radical acceptance, not deathism; given that one is in fact dying, whether one responds in a way that helps themselves and those around them seems like an important thing to ask about separately from the question of how much effort we should put into building a world where that is the case later and less frequently than the present.