“Is That a Lot?”: Rationality Workshop

Written By Chana Messinger

I’ve been thinking a lot about the epistemics of the Effective Altruism movement, and I wrote this workshop to address a set of ideas and concerns on my mind around groundedness in numbers, the ability to know what one believes and why, and the importance of knowing what would change your mind.

This is the core notes for this workshop, as I used them (cleaned up some). They were written to make sense to me personally; I hope they are also helpful to others. This also exists as a workflowy (with color coding and collapsible content) and a google doc.

I also wrote up my thoughts on the workshop, what I wanted to achieve, on which beats go where, my thinking behind each one, and my takeaways from feedback here.

  • Intro

    • Group Question: How many people here are familiar with the drowning child thought experiment?

      • Pick someone to explain

      • Group Question: How much do you buy this argument?

        • (For purposes of getting a sense of the group, checking in, etc.)

  • Discussion 1: 300/3000/3000

    • Ask about the numbers – what is your cached sense / what did Peter Singer say you can save a life for? Is that a lot?

      • Meta: For the purposes of this workshop, try to notice every time a number comes up and snap your fingers or tap your thigh and see if you can think “is that a lot”

    • How much does the convincingness of the argument rely on the numbers? What did you think when you originally heard it?

    • As it turns out, Peter Singer seems to have consistently said things about it being in the hundreds. It is more like in the thousands according to givewell (look up what number Givewell is currently using)

      • Evidence:

        • “But then it does occur to you that [saving the child] is going to ruin your most expensive shoes. You’ll be up for some hundreds of dollars to replace them and other clothes you might ruin. So, you think, why shouldn’t I just walk away and not have to go to the expense of replacing my shoes? ]

        • “We can all save lives of people, both children and adults, who would otherwise die, and we can do so at a very small cost to us: the cost of a new CD, a shirt or a night out at a restaurant or concert, can mean the difference between life and death to more than one person somewhere in the world – and overseas aid agencies like Oxfam overcome the problem of acting at a distance.”

    • So instead of say, $300, it’s more like $3000. What does that change? About our obligation? About our communication?

      • What if was 10x less or 10x more?

      • Scope sensitivity

      • And I mean, maybe it doesn’t matter! Holden makes this argument, and in broad strokes it doesn’t (see how this hits them)

      • Do you think it matters? Why?

        • Reasons it does matter (according to me, Chana Messinger)

          • We are truth tellers!

            • “Honestly, there really is a big difference to me if X is different by orders of magnitude. The U.S. federal minimum wage is currently $7.25 an hour. Payroll taxes are 7.5%, so take-home pay becomes $6.70 an hour. It takes 343 hours – two months, working full time – working a minimum wage job to earn the $2300 it takes your #1 charity to save a life. There’s a big difference between $200 and $2000, between one week of minimum wage work and two months of minimum wage work” Link

            • “Thought I was murdering 10x people than I was” – anonymous EA

          • But in comparison to other altruistic activities! We would want to know if something was way better

          • Also, people find this stuff a lot and destabilizating! It’s a lot as is, no need to make it more so by living in a world where every ice cream is a life.

          • Rigor matters and being easily criticized matters

          • Cause X, longtermism, etc.

  • Relevant Numbers

    • People in fact think it’s really cheap to save a life.

      • In fact, 40% of people believed the ‘most effective charities’ could save a life for under $10 – less than a hundredth of what we believe to be the true cost. 3% believed a life could be saved for under $1! A minority thought it would cost much more – 5% gave an answer over $1,000 – which pushed the average to around $500: https://80000hours.org/2017/05/most-people-report-believing-its-incredibly-cheap-to-save-lives-in-the-developing-world/

      • We have good evidence that the cost to save a life for typical charities is not anything like as low as $40. If this were true, then such charities spending $1 billion each year would be able to save 25 million lives – a sixth of all births globally, and more children than are actually dying of easily prevented diseases. There simply aren’t enough lives being saved in total for figures like that to add up.

      • “I gave this amount [Ed: $10 per life saved] because of advertisements I have seen on t.v. and the claims they make as to how much it cost to save a child’s life.”

      • “I remember seeing commercials saying for only $1 a day or $30 month you can save the life of a child. So I just guessed and picked half the amount.”

  • Discussion 2: You’re going to engage with a lot of EA content over time, and develop beliefs, and the hope is that they will be grounded. Numbers change a lot, if we’re going to be the kind of rigorous we claim to be, it needs to matter what the numbers are, and for our beliefs to change if the numbers do.

    • Do you feel like EAs throw around big numbers a lot? So much. If you don’t have your own sense of orders of magnitude, they’re just going to drag you around.

    • It’s important for trains to have brakes, for there to be numbers where you would get off the train.

    • “Is that a lot?” What would be big, what would be small?

    • Nate Soares says to shut up and multiply, but then the size of the numbers matters!

    • As Scott Alexander puts it, ‘when math tells you something weird, you at least consider trusting the math. If you’re allowed to just add on as many zeroes as it takes to justify your original intuition, you miss out on the entire movement.” A weird-seeming answer is a warning flag, rather than a stop sign: a thing to investigate rather than reject

  • Activity: So what’s a claim you care about? What numbers might you put on it? What is a number that, if a relevant variable were that value instead of the one you currently believe, would change your mind?

    • Analogy: Data blind analysis

    • It’s ok not to know, but worth having a sense, and a sense that you are entitled to a benchmark

    • (There may not be any one number for this, but you can generate multiple, or think about one that feels the most important even if on its own it won’t change your mind)

    • Is that a lot?

    • If you currently just have a guess at the number, and it’s important, can you find out in 5 minutes what the true number is? Cultivate curiosity and a willingness to Go Check.

      • There’s a lot of stuff you can just look up, quickly

      • And for numbers that are important to your worldview, maybe they bear doing some research into

      • You can get calibrated with various calibration tools to get better at thinking through numbers, as well as try Fermi problems

    • If the change-your-mind number is very far from what you currently think, your argument might be robust, and it means you can just track changes you should make when someone makes a pretty big claim. If it’s not that far, your argument is more sensitive to changes, so you might want to be more aware of what the current thinking on that number is. That seems good to know!

    • Maybe write down what you currently think that number is, look to see if there’s any change every month or year, and see how your beliefs do and don’t change.

    • To the extent you make this public, this is a way of being more accountable and demonstrating integrity. Brakes on your train.

      • “Well, in 2016, DeepMind revealed AlphaGo, and it was almost this canary in the coal mine, that Go was to me, that was sort of deep in my subconscious keeled over and died. That sort of activated me. I realized that for a long time, I’d said post tenure I would start working on AI. Then, with that, I realized that we couldn’t wait. I actually reached out to Nick Bostrom at the Future of Humanity Institute and began conversations and collaboration with them. It’s been exciting and lots of work to do that we’ve been busy with ever since.” - Allan Dafoe

    • It’s a way of doing the hard thinking in advance

    • I went around and asked people and liked doing that. So for instance someone said they cared about climate change and I asked what a mind-changing number change would be, and they said something like, if warming wasn’t coming until after 2075, I’d think there was something else I should work on.

    • This might feel bad (recall embracing the challenge!) but it’s also a way of dealing with epistemic helplessness – have a best guess, say what you’re doing with that, or say I don’t know, but if it went under this number I’d want to know, instead of having to wonder each time if it’s a thing you need to deal with. Hones in on what’s action relevant.

  • Example: Cows vs. Chickens

    • Explain the argument (without numbers)

      • The logic goes: the average cow is very big. The average chicken is very small. If you eat some number calories of meat per year, you can do that with way fewer cows than chickens, so eating beef reduces the number of suffering animals.

    • Discussion 3: Ask how many times more calories they think a cow is than a chicken, and on that basis to evaluate the argument. Ask around.

      • Likely people will bring up “but cows are smarter” and similar. Go with that! Lots of potentially relevant facts below, but more importantly, encourage incorporating that into their model, and encourage thinking about how big the number would have to be for that to matter (and note that there’s a table they can go fill out and put their own numbers in and see what direct suffering comes out)

The answer is 100x:

  • Potentially relevant facts

    • Demand for milk products is less elastic, but still significant – for every pound of milk you buy, you increase production by 0.56 pounds.

    • Per life, chicken produces around 1.9kg of meat, versus 212kg for cows

    • Due to elasticity, reducing consumption by 1kg only reduces production by ~0.7kg for both chickens and cows

    • The problem is still cost. Beyond Beef is $6.74/lb or $14.83/kg on Amazon (cross-check), versus just $8.80/kg for cow beef, or $10.93 for suffering and carbon offset cow beef. In other words, for the price of 1kg Beyond Beef, you could get a kg of cow beef, and use the remaining money to offset 6kg worth of meat. [2]

    • In that worldview, the real cost of 1kg chicken meat isn’t $8.55, it’s the 4 mosquito nets you could purchase for that same amount, with $0.55 leftover to eat rice and beans.

    • If we expect this to scale with neuron number, we find cows have 6x as many cortical neurons as chickens, and most people think of them as about 10x more morally valuable. If we massively round up and think of a cow as morally equivalent to 20 chickens, switching from an all-chicken diet to an all-beef diet saves 60 chicken-equivalents per year.

    • Cows are big, so raising one produces about 500 pounds of beef — and at the rate at which the average American eats beef, it takes about 8.5 years for one person to eat one cow. But chickens are much smaller, producing only a few pounds of meat per bird, with the average American eating about one whole chicken every two weeks. To put it another way, each year we eat about 23 chickens and just over one-tenth of one cow (and about a third of one pig).

    • Suppose that I stop eating chicken and switch entirely to beef. Now I am killing about 0.6 cows and 0 chickens, for a total of 0.6 animals killed. By this step alone, I have decreased the number of animals I am killing from 42.3/year to 0.6/year, a 98% improvement.

    • According to the National Agricultural Statistics Service, the average milk production per cow in the United States in 2006 was 19,951 lb/cow, which is equivalent to 2,320 gallons per cow per year.

    • And while a cow suffers and is slaughtered to produce around 500 pounds of meat, a chicken produces about four to five pounds of meat. So a switch from beef to chicken is actually a switch from a tough life for one cow to an awful life for around 100 chickens.

  • Citations

Google doc of Concrete Takeaways that I passed out at the end

The feedback form I’m currently using (I recommend splitting up most useful / least useful for better data, and only using the “probabilities of using this regularly” if you better define regularly and explain credences. There’s probably a better version of this question).

This content as a google doc

This content as a workflowy with color coding

Chana Messinger
Previous

Giving the “Is That a Lot?” Rationality Workshop