I’m going to share my answers. Please keep in mind that they might have been already tackled by other people elsewhere. In any case, those are the critiques I have so far.
Superficial references problem: The handbook almost never recommends books on the subjects (except those written by MacAskill, Ord, Singer, etc), but instead they tend to recommend blog posts, Wikipedia, other EA-aligned webpages, or, at best, philosophy papers. In my opinion, there could be recommendations of textbooks on cost-effectiveness analysis, cause prioritization, economics, ethics, statistics, cognitive biases, etc. Since webpages and standalone papers are not nearly as good as textbooks to learn a subject, I believe recommendations of books are definitely warranted, otherwise we can get the impression that all the theoretical background that EAs have are those shallow references.
The neglectedness problem: First, it’s not clear how to distinguish between these two scenarios: (1) The cause is unfairly neglected, that is, much more neglected than it ought to be, considering its scale and tractability; and (2) The cause is neglected because it’s really a bad cause to work for (due, for instance, to low scale or low tractability), in which case it being neglected is actually a sign that we shouldn’t work on it. In order to help us sort out what’s the underlying scenario, I think we should see whether other institutions/researchers have attempted to work on the issue in the past, and not just look at the absolute numbers of funding/researchers that are going to that cause in the present. I don’t remember seeing this historical analysis being done. And maybe we should employ other strategies besides this historical analysis to sort things out.
Second, there’s another shortcoming of just assessing neglectedness by looking at the amount of dollars being poured into a cause. People might be working to solve a problem and pouring lots of money into it using an inefficient method. For instance, suppose that we lived in a world where hundreds of billions of dollars were being spent on leafletting about the animal cause, and suppose that it is the case that leafletting is a very inefficient method to promote concrete changes to animal well-being. Then even though there are hundreds of billions of dollars being put into the Animal Cause, there would still be a low-hanging fruit, if we assume that, for instance, corporate campaigns are a hundred times more efficient than leafletting. So just looking at the sheer number of donated dollars to the Animal Welfare cause could be very misleading because even though it’s not “numerically” neglected, we’re not making use of the most effective methods.
Third, I’m not convinced that the curve of improvement as a function of funding/researching has a log shape (or any curve that implies diminishing marginal returns)
Fourth, even if it has a log shape, in order to infer that an additional person/dollar of funding would have a greater impact on cause A compared to cause B, we would need to know the parameters of the log curve for cause A and cause B, which we don’t. For example, check www.desmos.com/calculator/ohwiagg7zi. Here we have two causes with a log curve, but with different parameters (hence, different shapes), and we can see that even though cause red is receiving more funding than cause green, the marginal return of cause red is still higher than the marginal return of cause B, which makes comparisons between different cause with regards to neglectedness very hard, if not impossible.
Blindspots: By “blindspots”, I mean arguments that I’ve never seen being raised in the given discussions, though they seem to be crucial.
[The logic of the larder] blindspot in the Animal Welfare discussion: It’s not crystal clear what is the net value of the lives of each factory-farmed species. For instance, if some species have net positive lives, then interventions that aim to reduce the number of factory-farmed animals will cause a loss of total value. Another thing to consider is that, because of the crops to sustain factory-farmed animals, they have a negative impact on the number of wild animals, and if we consider that the lives of wild animals are worse than the lives of some factory-farmed animals, then abolishing factory farms will have this other source of disvalue as well, by creating lives whose quality is even worse.
[Intelligence restart] blindspot in the Extinction Risk discussion: If only humanity goes extinct, couldn’t some other species as intelligent as (or even more intelligent than) humans eventually evolve from other animals, say, from the surviving primates?
I’m going to share my answers. Please keep in mind that they might have been already tackled by other people elsewhere. In any case, those are the critiques I have so far.
Superficial references problem:
The handbook almost never recommends books on the subjects (except those written by MacAskill, Ord, Singer, etc), but instead they tend to recommend blog posts, Wikipedia, other EA-aligned webpages, or, at best, philosophy papers. In my opinion, there could be recommendations of textbooks on cost-effectiveness analysis, cause prioritization, economics, ethics, statistics, cognitive biases, etc. Since webpages and standalone papers are not nearly as good as textbooks to learn a subject, I believe recommendations of books are definitely warranted, otherwise we can get the impression that all the theoretical background that EAs have are those shallow references.
The neglectedness problem:
First, it’s not clear how to distinguish between these two scenarios: (1) The cause is unfairly neglected, that is, much more neglected than it ought to be, considering its scale and tractability; and (2) The cause is neglected because it’s really a bad cause to work for (due, for instance, to low scale or low tractability), in which case it being neglected is actually a sign that we shouldn’t work on it. In order to help us sort out what’s the underlying scenario, I think we should see whether other institutions/researchers have attempted to work on the issue in the past, and not just look at the absolute numbers of funding/researchers that are going to that cause in the present. I don’t remember seeing this historical analysis being done. And maybe we should employ other strategies besides this historical analysis to sort things out.
Second, there’s another shortcoming of just assessing neglectedness by looking at the amount of dollars being poured into a cause. People might be working to solve a problem and pouring lots of money into it using an inefficient method. For instance, suppose that we lived in a world where hundreds of billions of dollars were being spent on leafletting about the animal cause, and suppose that it is the case that leafletting is a very inefficient method to promote concrete changes to animal well-being. Then even though there are hundreds of billions of dollars being put into the Animal Cause, there would still be a low-hanging fruit, if we assume that, for instance, corporate campaigns are a hundred times more efficient than leafletting. So just looking at the sheer number of donated dollars to the Animal Welfare cause could be very misleading because even though it’s not “numerically” neglected, we’re not making use of the most effective methods.
Third, I’m not convinced that the curve of improvement as a function of funding/researching has a log shape (or any curve that implies diminishing marginal returns)
Fourth, even if it has a log shape, in order to infer that an additional person/dollar of funding would have a greater impact on cause A compared to cause B, we would need to know the parameters of the log curve for cause A and cause B, which we don’t. For example, check www.desmos.com/calculator/ohwiagg7zi. Here we have two causes with a log curve, but with different parameters (hence, different shapes), and we can see that even though cause red is receiving more funding than cause green, the marginal return of cause red is still higher than the marginal return of cause B, which makes comparisons between different cause with regards to neglectedness very hard, if not impossible.
Blindspots: By “blindspots”, I mean arguments that I’ve never seen being raised in the given discussions, though they seem to be crucial.
[The logic of the larder] blindspot in the Animal Welfare discussion: It’s not crystal clear what is the net value of the lives of each factory-farmed species. For instance, if some species have net positive lives, then interventions that aim to reduce the number of factory-farmed animals will cause a loss of total value. Another thing to consider is that, because of the crops to sustain factory-farmed animals, they have a negative impact on the number of wild animals, and if we consider that the lives of wild animals are worse than the lives of some factory-farmed animals, then abolishing factory farms will have this other source of disvalue as well, by creating lives whose quality is even worse.
[Intelligence restart] blindspot in the Extinction Risk discussion: If only humanity goes extinct, couldn’t some other species as intelligent as (or even more intelligent than) humans eventually evolve from other animals, say, from the surviving primates?