On a personal level, I share some of your sadness at 80k's change in focus. I used to listen to every single episode of the 80k podcast and recommend it to anyone who asked, and I just don't get the same level of enjoyment out of it without the previous level of content diversity. But I can't fault 80k for their overall change in focus. AI risks that looked speculative as of a few years ago are seeming increasingly close to real, whether that's loss of control (see the hugging face swarm) or AI-enabled cyber risk (see the recent discovery of vulnerabilities in PHP). With that in mind, I think that several of the problem profiles you highlighted as "esoteric and strange" really aren't that out-there, and I expect that many visitors to the 80k website who have seen some of the latest AI news would agree.

Moreover, EA is about trying to do good with evidence and reason, and that means it needs to involve a balance between identifying the most effective ways to do good and actually going and doing those things. So while I agree that it's a loss to no longer have as good a cause-neutral EA introduction and to lose a common intro EA funnel, I think that can reasonably be counterbalanced by having a better resource for getting people to work on addressing pressing problems related to AI. Not every EA org needs to be cause-neutral, and it's reasonable for organizations to shift their role in response to changes in circumstances.

Thanks for writing this up! It's rare to get an inside look at how funders handle the practical challenges of allocation, and I think you had a lot of interesting ideas. 

However, the various approaches you discussed seem odd to me in that they all involve taking the cost-effectiveness of the intervention and probability of its successful implementation and converting them into some kind of score, then allocating funding proportionately to that score. That approach seems like it misses something fundamental, which is that an EV-maximizing grant portfolio should be one that equalizes the marginal cost-effectiveness (including value of information) of the last dollar given to each recipient (note that there's some further complexity if you have flexibility in how much funding to give this year vs. hold onto for future years, since one should also want to equalize the marginal cost-effectiveness of grant dollars over time). Starting from that goal, it seems clear that it would be worth scoring organizations on how quickly cost-effectiveness would be likely to fall off, based on total addressable problem size, management capacity etc. That's not to say that one can in practice compute exact marginal cost-effectiveness curves and perfectly solve this optimization problem, but I do think it's worth bringing that marginal mindset to evaluating each grant.

I'd expect the portfolio that resulted from thinking hard about equalizing marginal cost-effectiveness to look something like:

• Give a lot to the organizations where the first dollar has high expected cost-effectiveness, filling them until they hit serious diminishing returns. (e.g. H, G, J)
• Give grants just large enough to get significant information value to organizations that have high impact conditional on successful implementation but low probability of success, since the value of information there is very high (e.g. B)
• Give nothing to organizations that have low impact conditional on successful implementation and low probability of success, since the value of information there is low (e.g. C, K)
• Give the remainder to organizations in the middle, or do something else with it (hold it for future years, give to an outside option that can absorb lots of dollars at a high cost-effectiveness floor like GiveDirectly)

This looks a decent amount like edge, but it's skewed somewhat away from the middle tier. For example, I'm skeptical that edge's $1.9m grant to F makes a lot of sense, given that F's EV is much lower than the top tier of organizations and the information value if it works is still low. I'd expect the recommendations to differ more from edge in cases where organizations had very different capacities to absorb funds, which this example doesn't really capture.

Thanks!

I was curious about your comment that "we intend to create cost-effective impact for farmed animals by making the grants that other funders cannot - small grants delivered early and fast." From this initial description, it doesn't seem obvious to me that you're targeting a different niche than what ACE Movement Grants and AWF are currently covering. Could you say more about how you see BluePerch differing from these existing orgs? (Though I will note that it's not a bad thing to have some overlap here, more grantmaking capacity for effective animal advocacy is great!) 

I think I concur. It seems like a bad call to put someone who committed massive financial crimes into a grantmaking platform development/research role, especially one in the EA/AI safety community. I don't disagree that redemption is possible and that people can change, but I think the conditions for hiring someone in these kinds of cases (e.g. time since the crime occurred and weight of evidence that the person has changed) need to be a function of the sensitivity of the role and the magnitude of the crime. And I don't think that bar is anywhere close to having been met here. 

Super cool! 

Thanks for the post! I was surprised to see how weak the academically-published evidence is regarding shrimp stunning effectiveness, and agree that it would be valuable on current margins for effective animal advocates to invest more in research. 

For what it's worth, I read the Tesco-Hilton case study more optimistically than you do. You quote it as:

The results are a bit vaguely presented, but with at least one machine setting “a significant proportion” of shrimp did not “show signs of recovery” within 10 minutes.

But the full quote is

A significant proportion of prawns were in irrecoverable stun (stun-kill) on exit of the stunner as evidenced by transfer back to a controlled aqueous environment where none showed signs of recovery within the monitoring period (10 minutes plus handling time of 32 seconds)

(emphasis mine). That's not to say that we should rely on a single vague industry report when the academic evidence is conflicting, but I think the report does provide some evidence. 

Thanks for engaging with the post! You made a lot of different points, so I'll do my best to separate them out and consider them one-by-one: 

(1)

• I'm not making an argument for quietism. Saying that we have an obligation to follow the law is compatible with having obligations (even extraordinarily strong ones) to use non-illegal means to combat injustice (e.g. by advocating for changes to laws).
• It's a genuinely interesting point that many of our laws are inherited traditions, rather than the direct product of the democratic process. However, I don't think that's a strong argument in this specific case. The US has had true universal suffrage for more than 60 years, and in that time Congress and state legislative bodies have passed many laws related to the treatment of animals and the criminality of trespassing. Under any reasonable interpretation of democratic legitimacy, a democratically-elected legislative body specifically dealing with an issue and choosing to pass laws that accept the underlying common law principles and add specific penalties, related rules etc., should confer it.
• I don't disagree that a reasonable contractualist would think that there are cases where it would be justified to break an unjust law. The core question is whether the required conditions hold in this case. Democratic legitimacy is one important part of that, since reasonable contractualists generally would give some weight to whether laws resulted from a just process. A point I didn't make in the OP, but I think is relevant here, is that even if you disagree about the democratic legitimacy argument, I think the specific nature of the lawbreaking here falls outside many notions of justifiable civil disobedience. That's because the Ridglan rescues involved breaking a law to achieve a non-symbolic end (rescuing the dogs), not merely symbolically challenging a law by breaking it.

(2)

• I think you're moving between a couple different notions of universalizability here. It's true literally everyone breaking and entering in service of moral aims is a far-fetched idea. But it's still coherent to ask whether a tactic would have positive or negative effects if commonly used across social movements. Democratic societies can and have experienced periods of widespread civil unrest.
• I agree that a similar argument could have been deployed against rebellion by enslaved people, but I think the analogy is weak because of the specifics. Slave rebellions occurred in a society where the affected population was excluded from political participation, and the principles justifying them were self-limiting in the way I described in the OP. The current case is different: the affected parties (animals) can't be enfranchised, but the human population that cares about animal welfare has full political participation, and the legal channels for advancing animal welfare are open and have produced incremental gains over recent decades (most notably the transition of nearly half of the US egg supply to cage free). The bar for lawbreaking is plausibly higher when those channels are more responsive than when they're closed.

(3) 

• I think you're being too quick to dismiss property as being something that can drive moral obligation. There are clearly many cases where we are obligated to not destroy or interfere with others' property, as is obvious in cases where vulnerable groups' property rights are infringed. The way I'd think about this is that obligations to protect property are stronger the more just a society's system of property rights is. In a more just society, property destruction not only weakens otherwise-good norms, but is also more likely to be the result of a miscalculation: if the base rate of unjust property ownership is lower, then any given case of someone believing property destruction is justified is more likely to be wrong. So both the rule-following considerations and the act-utility considerations point toward higher property protection in more just societies.

(4)

• I disagree that the priority should be supporting one another within a social movement; the priority should be trying to do the most good. Donor considerations reasonably impact that calculation, both because money is a necessary ingredient for advocacy and because donor preferences may reflect genuine moral views that are worth considering. But I do also agree that it can be worth trying to convince donors of an approach rather than just deferring to their preexisting preferences.

I see what you're getting at here. But if we agree that the externalities of crime aren't internalized, then I think we're just back in the position of the original post. You think the act utilitarian calculus checks you, I'm both skeptical that it does and think that there are non-act-utilitarian reasons why we ought to avoid lawbreaking. 

On (1): You say: 

Many people believe moral-ish things strongly and don't find property destruction immoral, but far far fewer actually destroy the property of those they think are doing something immoral. This seems like good evidence that the expected punishment (including via informal mechanisms) is not too light.

I think that this is at best weak evidence. Activists' decisions of whether or not to commit crimes are surely influenced by norms, not just the expected intensity of punishment. The recent history of climate activism in the UK is a good example. As far as I can tell, nothing changed about UK law to cause the rapid rise of high-profile lawbreaking by Extinction Rebellion and then Just Stop Oil in the 2018-2023 timeframe. The UK government did in the end stop the activists through increased legal penalties (going from typically no prison time for nonviolent lawbreaking when motivated by ethical concerns to 4+ year prison sentences becoming common for the more serious cases). But something other than threat of prison time was keeping climate activists from using these tactics in the early-to-mid 2010s. 

On (2): 

I agree that a Pigouvian tax doesn't require restitution (as I indicated with including "improve welfare elsewhere" as something that can be done with the tax revenue). But the classical formulation (in which the optimal tax rate fully eliminates deadweight losses) does require that a dollar of consumer/producer surplus and a dollar of tax revenue produce the same social welfare. If a dollar of tax revenue produces less social welfare, then the deadweight loss cannot be eliminated. 

To make this more concrete, I want to dig into an example based on your comment about driving in a world with a carbon tax. Consider taking a long trip by car rather than train under 3 different taxation schemes. Let's assume you value the convenience of the car over the train at $101, the social cost of your carbon emissions is $100, and that all consumers in this world have identical marginal utilities of money.

World A: No carbon tax. You take the trip, gaining benefits you value at +$101 while causing social costs of -$100. In this world, we might say that you've done the right thing by driving (since this maximizes utility overall), but that for fairness reasons you might be obligated to donate some money to others, since your utility-maximizing decision also acted as a transfer from others to you. 

World B: Carbon tax of $100 on the trip, returned as an equal dividend to all people. You take the trip, gaining net benefits after the tax of $1. The rest of society ends up net neutral (though there might still be particular winners and losers). In this world, you've done the right thing by driving, and have no further obligations. 

World C: Carbon tax of $100 on the trip, which the government will use to buy $100 of consumer goods and dump them down an old mineshaft. You take the trip, gaining net benefits after the tax of $1. The rest of society still experiences the social costs of -$100, which the tax doesn't do anything to reduce. In this world, you've clearly done the wrong thing by driving, since you caused a net utility loss of $99.

My claim is that doing crimes is similar to deciding to drive in World C. The "tax" on crime is imprisoning the criminal, which causes them to pay large costs in terms of their lost freedom and ability to work and doesn't do anything to benefit society. And in fact it's worse than World C, since the rest of society needs to pay the additional costs of arresting, prosecuting, and jailing them. So I think the Pigouvian tax analogy does not hold here, and it's wrong to think that the harms of crime are properly internalized.