Or, to change tack: the operating budget of the LessWrong website has historically been ~$600k, and this budget is artificially low because the site has paid extremely below-market salaries. Adjusting for the market value of the labor, the cost is more like $1M/year, or $2,700/day. If I assume LessWrong generates more value than the cost required to run it, I estimate that the site provides at least $2,700/day in value, probably a good deal more.
I think this estimate is mistaken because it ignores marginalism: basically, the cost of disabling LW for a year is much larger than 365 * the cost of disabling LW for only a day. The same goes for disabling the whole website vs. disabling only the frontpage.
(Sorry for adding salt to hurt feelings; posting because impact evaluation of longtermism projects is important.)
I will note that the curves here are non-intuitive, for example I suspect that 2 weeks down is less than 2x as bad as 1 week down, because the first week is enough to mess everyone’s “check LessWrong” habit and cause long-term losses to LWers contributions, and so most of the value is already lost before the 2nd week down.
(Note: I don’t know if I stand by the 1 vs 2 weeks in particular, those are example numbers.)
If someone loses a LessWrong habit, the impact could be positive or negative, depending on what they contributed to LessWrong, and what they replace the habit with. Most communities contain some fraction of members who would be better placed elsewhere, not because the community isn’t providing value, but because it’s now providing less value, or because new opportunities have arisen. Those people are disproportionately likely to drift away during a disruption. People for whom LessWrong is the best marginal use of their time will stick with LessWrong.
Various non-essential services in the US are unavailable during Thanksgiving and Christmas, and this seems to be viewed positively overall. Overall, I don’t know what direction the habit-breaking effect would go. For the 2020 Petrov Day exercise I’d expect it to be small and net positive.
The estimate also misses the mark because the stakes were not disabling LessWrong for a day. When the button is pressed, only the front page goes down. RSS feed is unaffected, search engine results are unaffected, etc. Eg, see Slider, Measure.
A Fermi estimate: given the investment of time and money in LessWrong, I estimate that it creates $10M/yr in rationalist value. Not bad. But for a one day front page outage I apply a 90% discount due to diminishing returns and a further 90% due to only the front page going down. That gets me to $300 of value lost. Not nothing, but a different scale.
If I didn’t apply those discounts, and genuinely thought that $30,000 of rationalist value was at stake, I’d think that the Petrov Day ritual was a terrible idea! Given the 33% observed chance of launching “nukes”, that’s only justified if the ritual creates more than $10,000 of value. The claim is that the ritual increases overall trust, so this increased trust has to lead to at least a 0.1% increase in productivity. That’s a large effect size for (a bit more than) sending a bulk email that 100 people opened.
I appreciate the Fermi. It’s a fair point that only the frontpage is taken down.
I think that a well-designed ritual ought to create $10,000 of value if it’s to justify the amount of time spent on it (1-2 person-weeks). Or at least if I didn’t think it would create that much value, I wouldn’t choose to do it.
I don’t see why sending a 100 emails and announcing that generally couldn’t have a very large effect size. All depends on the email.
I think this estimate is mistaken because it ignores marginalism: basically, the cost of disabling LW for a year is much larger than 365 * the cost of disabling LW for only a day. The same goes for disabling the whole website vs. disabling only the frontpage.
(Sorry for adding salt to hurt feelings; posting because impact evaluation of longtermism projects is important.)
I will note that the curves here are non-intuitive, for example I suspect that 2 weeks down is less than 2x as bad as 1 week down, because the first week is enough to mess everyone’s “check LessWrong” habit and cause long-term losses to LWers contributions, and so most of the value is already lost before the 2nd week down.
(Note: I don’t know if I stand by the 1 vs 2 weeks in particular, those are example numbers.)
If someone loses a LessWrong habit, the impact could be positive or negative, depending on what they contributed to LessWrong, and what they replace the habit with. Most communities contain some fraction of members who would be better placed elsewhere, not because the community isn’t providing value, but because it’s now providing less value, or because new opportunities have arisen. Those people are disproportionately likely to drift away during a disruption. People for whom LessWrong is the best marginal use of their time will stick with LessWrong.
Relevant study: The Benefits of Forced Experimentation: Striking Evidence from the London Underground Network
Various non-essential services in the US are unavailable during Thanksgiving and Christmas, and this seems to be viewed positively overall. Overall, I don’t know what direction the habit-breaking effect would go. For the 2020 Petrov Day exercise I’d expect it to be small and net positive.
The estimate also misses the mark because the stakes were not disabling LessWrong for a day. When the button is pressed, only the front page goes down. RSS feed is unaffected, search engine results are unaffected, etc. Eg, see Slider, Measure.
A Fermi estimate: given the investment of time and money in LessWrong, I estimate that it creates $10M/yr in rationalist value. Not bad. But for a one day front page outage I apply a 90% discount due to diminishing returns and a further 90% due to only the front page going down. That gets me to $300 of value lost. Not nothing, but a different scale.
If I didn’t apply those discounts, and genuinely thought that $30,000 of rationalist value was at stake, I’d think that the Petrov Day ritual was a terrible idea! Given the 33% observed chance of launching “nukes”, that’s only justified if the ritual creates more than $10,000 of value. The claim is that the ritual increases overall trust, so this increased trust has to lead to at least a 0.1% increase in productivity. That’s a large effect size for (a bit more than) sending a bulk email that 100 people opened.
I appreciate the Fermi. It’s a fair point that only the frontpage is taken down.
I think that a well-designed ritual ought to create $10,000 of value if it’s to justify the amount of time spent on it (1-2 person-weeks). Or at least if I didn’t think it would create that much value, I wouldn’t choose to do it.
I don’t see why sending a 100 emails and announcing that generally couldn’t have a very large effect size. All depends on the email.