The great calculator
I’ve made a calculator which I’m looking to get some feedback on. I don’t think I’ve ever seen anything similar.
It’s purposefully simplified, yet I still think it’s quite illuminating.
I’m also being purposefully vague, as the impact is in not knowing what the calculator is about before using it, as that will bias the results. I will put a more detailed description in the comments at a later time.
This calculator was inspired by a conversation with Brenton Mayer at the EA Global conference in Melbourne, 2015.
I like this idea, and I think an improved version could be helpful in clarifying people’s thinking about the importance of reducing animal suffering. But I see a few substantial problems with the existing spreadsheet, and I don’t think the results can be considered meaningful until these problems are addressed.
A person who avoids eating animals is reducing—not increasing—animal life-years, by preventing factory farm animals from being born. The same is true for most THL’s vegan outreach interventions, assuming such interventions are effective. That being said, if a factory farmed animal has a life that is not worth living (and I think that describes most factory farmed animals, though not all) then reducing the life-years of such animals is a good thing. Or alternatively, if one’s goal is to reduce the act of killing animals, then reducing the life-years of such animals may be an acceptable cost. But since these goals are very different from saving the lives of animals in the same way that donating to AMF might save the life of a child, they cannot be considered the same thing, and are not comparable without making further assumptions. This seems like a tricky problem to solve in your spreadsheet.
THL mostly conducts vegan outreach, and the evidence that vegan outreach is effective at reducing consumption of animal products is weak. The evidence for specific cost effective figures like $0.60 for a 1-year reduction in animal suffering is even weaker. I do think the evidence for THL’s cage-free campaigns is substantially better (though still very far from GiveWell-quality evidence). Given Open Philanthropy’s recent grant to THL, however, it appears THL’s cage-free program (at least in the US) has no room for funding for the time being. So I take issue with confident statements that one can reduce 1-year of animal suffering with a $0.60 donation. I think such statements should include caveats about the very weak evidence for these cost effectiveness figures.
You ask people to estimate the value of a pig’s well-being relative to a human’s well-being, but you’re presumably talking about eating animals in general when you say that going vegetarian saves 100 animal lives, or when you say donating to THL reduces one year of animal suffering for $0.60. But in terms of numbers, the vast majority of those 100 animal lives a person eats is going to be chicken and seafood, and I think most people would not give as much weight to the suffering of one chicken, fish, shrimp, etc. as they would give to the suffering of one pig. If you use chicken instead of pig in the spreadsheet, I think that would be more reasonable.
You suggest that the person should become vegetarian, but I personally don’t think this is an efficient compromise. In general, the consumption of eggs in the US causes a considerable amount of suffering, I think far more than beef. (I personally think beef cattle have lives worth living.) It certainly causes considerably more animals killed than the consumption of beef (though one might give a higher weight to killing a cow vs. a baby chick or hen). So an effective altruist with the goal of reducing animal suffering and/or killing without becoming vegan should probably limit their consumption of animal products to dairy and beef instead of dairy and eggs.
Avi
Thanks for that Avi. As I’ve said elsewhere, this was a very quick proof of concept to get a feel for whether developing this into a website or something similar would be useful—which is why there are a lot of assumptions.
You’re right. I oversimplified the language and maths to keep the message clean and simple (and in my opinion more impactful), but this seems to be proving rather unpopular.
More reasonable, but less impactful, which should be weighed carefully.
This is cool. There needs to be more of these types of things.
That said, I don’t agree with much of the approach of this.
Why tell people to ‘enter the first answer that comes to you’? These are complicated questions and necessitate consideration.
These are not things people are used to thinking about, so that biases the results. I’d consider alternative questions.
Is there some reason that you can’t provide any context at all? I can think of many ways to provide partial context, so that people aren’t going into this blind, without ‘giving away’ what it does.
All very fair points Josh. Like I said, early work in progress, which is why I’m here.
Agreed.
Agreed—any suggestions?
Any suggestions?
Agreed.
I think a polished version of this calculator could be an effective way to illustrate the importance of animal suffering.
That said, all of your assumptions differ from ACE’s (see here, here). Why is that?
Also, the spreadsheet assumes that all animal lives are pig lives.
Given that the average pig is more intelligent than the average wild shellfish, farmed shellfish, wild fish, farmed fish, or chicken (all more impacted by animal advocacy according to ACE), your calculator will be unfairly more striking to meat eaters who believe consciousness and intelligence are correlated.
Thanks for your feedback Mac.
My assumptions differ as I used the stated average cost to reduce a year of suffering from Doing Good Better, which was, from memory, 60 cents.
As I’ve said elsewhere, this was a very quick proof of concept to get a feel for whether developing this into a website or something similar would be useful—which is why there are a lot of assumptions. But even so, with an appropriate footnote, I’m not convinced that having a lot of assumptions is a bad thing. The object is to reduce suffering, not to have a ‘down to the wire’ rigorous calculator. Given that, I’m not even convinced that biasing the results, e.g. by making it more striking, would be bad either, again with the appropriate footer. Such techniques are extremely common in the advertising space, and for good reason—they work.
As an aside, you’re right about the pigs, but the ACE calculator shows 180 chickens are also spared a life of suffering by best guess estimate, at a cost of 28 cents a chicken. As chickens are also quite intelligent, might you suggest switching the calculator to chickens?
Interesting idea. Main thing that bothers me is the conflation between “animal lives” and “animal life-years”. It’s standard in development economics to use life-years as the relevant measures, not total amount of deaths-averted. So I think using life-years as the measure would result in more accurate results (e.g. ask how much a person is willing to pay to save 15 years of pig-life).
I completely agree, but see my other comments about the relationship of accuracy/rigour vs impact.
I think you meant “avert 15 years of pig-life in industrial agriculture”
Nice—how about adding a GCR component? For instance, you could ask how much people value future lives as a fraction of current lives. Then you could say even if the only GCR were asteroid impact and even if you only thought humans would exist as long as the average mammal species, if you didn’t care when people were born, you could still save life years at only $2.5.
Good idea, I’ve had a similar discussion with someone else about this. I think that would be a good idea for a separate calculator more targeted at EAs, but since this one is targeted at those who are non-EA aligned, I don’t think that’s a good idea for the following reasons:
It would complicate the calculator, not the point of being too complicated to be accurate, but to be too complicated to resonate or be meaningful for someone from the vast non-EA population of people who don’t find such rigour to be very compelling.
I think it would be easier to get people thinking about their morals and effectiveness in present animal terms (human and non-human) first than to go straight for x-risk.
There could even be a follow-up calculator after this simple one (e.g. “Feeling compelled? Click here for an even more shocking calculator”).
Really cool Michael. Totally worth making this a webpage. 1) Is it a pig or a dog? The type of animal would elicit different answers for many people. 2) I think it would be good to have an explanation spelling out how the final numbers are calculated 3) Is there a reason for “if you were the only one who could do so?” I take it is because only we can save the animals we eat? If so, I get the idea. But I think that the statement makes it more emotional. Perhaps that is the idea. I felt that the statement compelled me to put a larger number in than I would have otherwise.
Thanks! That’s the intention eventually!
1) Good point. I had originally set it as dog as I thought that would elicit a stronger response, then I changed it to pig to be more in line with what animals people tend to eat. I thought about a question like, “Do you agree that a pig can suffer as much as a dog?” but this starts to get complex.
2) Good point, I can easily add that to the text.
3) I think that’s what I was going for. I wanted to get people to think about what they’d personally be willing to sacrifice in a very ‘child in the pond-like’ scenario, and they can’t divert moral or practical responsibility. Obviously that will bias the user to inputting a larger number there, but that’s my point, for the most part.
This looks really good! As my goal is to improve our ability to communicate to a broad audience, I’d suggest visually representing the outcome in the “number of pigs” or something like that, which will make it more impactful.
Thanks for the feedback everyone. Lots of recurring themes, so I’ll address them partly here.
The main point is this; the end market is not Effective Altruists. I don’t think it’s very likely at all that adding too much complexity for the sake of accuracy, at least on the front end, will result in any meaningful reduction in animal suffering. The point is not to be deceitful or to bias people, but simply to maximise the reduction in animal suffering.
As someone said at the EA Global 2015 conference in Melbourne, “Sometimes the best way to be a utilitarian is to pretend to not be a utilitarian”, which I loosely take to mean that we should sometimes drop our perceived moral or analytical rigour in order to actually do more good.
Perhaps there could be two versions; one which is completely rigorous, contains elements of x-risk (as some people have suggested) and is targeted at existing EAs, while the other is targeted at the broader public.
On a related note, I’m yet to do the calculation but I’m of the mind that current estimates on animal welfare charities are actually underestimates as they don’t factor in the long-run benefits of reducing the proportion of humanity that relies on subjecting animals to suffering. The earlier a society that doesn’t inflict suffering on animals is brought about, the fewer future animals will suffer. I find this tends not to be even mentioned when people compare animal welfare orgs to x-risk orgs.
But I’m very open to continuing this discussion. As I’ve said these are early days for what was an idea I wanted to get in the public space.
I think you’re conflating a couple of different dimensions: degree of complexity, and degree of rigour.
These two are linked: there are some aspects that it’s hard to be rigourous about without a certain level of complexity. But it can also be more work to make a more complex model rigourous, because you need to be careful about more different moving parts.
I think for a calculator like this you should be aiming for low complexity and high rigour. Adding more questions or complicated arguments could put people off. But making elementary mistakes or sleights-of-hand in conversion makes it easier to attack (and people will try to attack it) and dismiss. So keep the number of questions small—addressing existential risk definitely looks like a mistake to me—but try to make them the most appropriate ones, and keep the language precise. This recent post on depicting poverty and Josh’s comment there have some good discussion of what kind of language will avoid pushback.
Great comment, you’ve convinced me. Thanks for the link as well, it looks interesting.
I think the idea of this calculator is pretty good, but the implementation pulls some pretty questionable moves which in aggregate make it quite misleading (perhaps multiple orders of magnitude).
Some of the concerns (robustness of cost-effectiveness estimate; switch between pig lives and all animal lives; distinction between saving lives and averting them) have been highlighted by others. Two other serious ones:
1) Adding “once off situation, if you were the only one who could do so” to the question about saving a human’s life. These conditions don’t apply to the animal lives in consideration later. There is a case that that shouldn’t make a difference, but I think this is controversial, and to the extent that it shouldn’t make a difference I think you shouldn’t include the condition when asking about human lives.
2) Switching between trade-off rates for time-suffering between humans and pigs and trade-off rates for saving lives. Since humans live many times longer (very roughly one hundred times longer) than pigs who are raised for meat, just on suffering terms you might get quite a large factor of divergence when you do this conversion. Additionally, when thinking about saving lives instrumental effects may be more important to our judgements than when comparing suffering, and this could further skew the conversion. You could correct for this by asking about the trade-offs between lives directly, or asking about trade-offs between suffering and saving/averting lives.
Thanks for your comments, see my other responses, particularly around the question of rigour vs. impact.
Question “How much would you be willing to donate to save a human’s life?” makes no sense to me. It all depends on how old the human is, how happy/miserable his future life is going to be and… how much meat he eats.
By saving non-vegan’s life, you kill all the animals he is going to eat (well, in reality it’s more complicated, with market elasticities, etc.). If we are going to be perfectly rational, we should first calculate whether we even want to save the life.
Thanks for your comments, see my other responses, particularly around the question of rigour vs. impact.
I’ve performed a similar calculation based on fine-grained data for different species and animal years (courtesy of ACE) in my cause selection post.
I really like this, very worthwhile, and needs to be shared a lot once fully formed!