Super nice post. Impressed by the 'translation into EA language' from the original post. It's also a skill I'm working on.
I'd like to ask a little more about your point in "2. Fund rigorous evaluations of participatory allocation" How does this process respond to participatory grantmaking inherently having a lower statistical verifiability, given that participatory grantmaking is generally much less uniform and traceable using quantitative techniques? There's no good counterfactual for Extinction Rebellion (funded by Guerilla Foundation), and FundAction's grantees are tiny and extremely diverse. I've seen granters have to give up on statistics in these cases and lean into other ways of knowing e.g. relationships, trust and feelings dreaded by all EAs;) I think that a vast amount of time and resources will be needed to get a large enough sample size, and I'm not sure that social movement groups will be too happy about that.
A minor maths point but it will be important if you start running these calculations in real contexts.
"Maybe they took a paycut equal to EUR 100 for this life saved, then their "share" of this life in this case would be 100/3000 = 3%."
Actually, here the real cost of the life saved is 3100 EUR, not 3000. If it helps, you can imagine the employee being paid the competitive salary, then donating back the 100 EUR.
Thanks very much for all these great hypotheses. I think there is a reasonable chance that these effects will be true, but also that we don't have strong evidence for any of them yet. I've split your comment into the following hypotheses, and tests that can be run to determine whether they are true:
Hypothesis 1: Spread rate will be similar to GWWC 10% pledge, at 10,000 members over 15 years of activity. GWWC currently have 15 employees, with a time averaged number of c. 10 employees. This puts the spread rate per employee at around 0.2 promisers per day, compared to the model's current estimate of 3. This would lower the QALYs per dollar from 915 to 123. Test for hypothesis 1: monitor the spread rate per employee
Hypothesis 2: People gaming the system by becoming aware of the promise, but only taking the promise once they are in need, will be a significant effect. Test A for hypothesis 2: Employees record interactions with potential promisers and indicate what proportion of them seem likely to engage in this behaviour Test B for hypothesis 2: Monitor the spread rate by employees and by promisers directly, ignoring this intermediate variable
Hypothesis 3: There will be an overwhelming number of promisers without access to food, water, peace and shelter, which will lead to a high defection rate (above the 0.5% per day predicted in the model) Test A for hypothesis 3: Record the defection rate from the promise by running trials Test B for hypothesis 3: Observe the ratio of promisers who are able to give and promisers who are making requests for support
If you or anybody else would be interested in supporting the running any of these tests, please let me know.
Could you suggest an alternative title? Maybe you're thinking something like: "we should set up a pilot charity to formalise the promise - predicted to save 915 QALYs per dollar"?
There is no website as far as I have seen. It is not a formal organisation, but rather a text. There is, because of this, no centralised register of who has taken the promise, though I'm sure that would be welcomed, and could be one of the activities that the created charity does. So far, the promise has been working via word of mouth - if you need help, you ask Promisers that you know, and they ask Promisers that they know. Cool about the notion page. Maybe you could link it?
Ah thanks for pointing out my mistake! And yes, I read this paragraph in the article, but still couldn't work out how they could provide such a precise range
p.s. I'm Sophie Davison on LinkedIn
Super nice post. Impressed by the 'translation into EA language' from the original post. It's also a skill I'm working on.
I'd like to ask a little more about your point in "2. Fund rigorous evaluations of participatory allocation"
How does this process respond to participatory grantmaking inherently having a lower statistical verifiability, given that participatory grantmaking is generally much less uniform and traceable using quantitative techniques? There's no good counterfactual for Extinction Rebellion (funded by Guerilla Foundation), and FundAction's grantees are tiny and extremely diverse. I've seen granters have to give up on statistics in these cases and lean into other ways of knowing e.g. relationships, trust and feelings dreaded by all EAs;)
I think that a vast amount of time and resources will be needed to get a large enough sample size, and I'm not sure that social movement groups will be too happy about that.
A minor maths point but it will be important if you start running these calculations in real contexts.
"Maybe they took a paycut equal to EUR 100 for this life saved, then their "share" of this life in this case would be 100/3000 = 3%."
Actually, here the real cost of the life saved is 3100 EUR, not 3000. If it helps, you can imagine the employee being paid the competitive salary, then donating back the 100 EUR.
I think Malawi's colonial legacy holds a lot of answers to this question: https://www.lawsociety.ie/gazette/in-depth/effects-of-colonialism-on-malawi
Thanks very much for all these great hypotheses. I think there is a reasonable chance that these effects will be true, but also that we don't have strong evidence for any of them yet. I've split your comment into the following hypotheses, and tests that can be run to determine whether they are true:
Hypothesis 1: Spread rate will be similar to GWWC 10% pledge, at 10,000 members over 15 years of activity. GWWC currently have 15 employees, with a time averaged number of c. 10 employees. This puts the spread rate per employee at around 0.2 promisers per day, compared to the model's current estimate of 3. This would lower the QALYs per dollar from 915 to 123.
Test for hypothesis 1: monitor the spread rate per employee
Hypothesis 2: People gaming the system by becoming aware of the promise, but only taking the promise once they are in need, will be a significant effect.
Test A for hypothesis 2: Employees record interactions with potential promisers and indicate what proportion of them seem likely to engage in this behaviour
Test B for hypothesis 2: Monitor the spread rate by employees and by promisers directly, ignoring this intermediate variable
Hypothesis 3: There will be an overwhelming number of promisers without access to food, water, peace and shelter, which will lead to a high defection rate (above the 0.5% per day predicted in the model)
Test A for hypothesis 3: Record the defection rate from the promise by running trials
Test B for hypothesis 3: Observe the ratio of promisers who are able to give and promisers who are making requests for support
If you or anybody else would be interested in supporting the running any of these tests, please let me know.
Could you suggest an alternative title? Maybe you're thinking something like: "we should set up a pilot charity to formalise the promise - predicted to save 915 QALYs per dollar"?
There is no website as far as I have seen. It is not a formal organisation, but rather a text. There is, because of this, no centralised register of who has taken the promise, though I'm sure that would be welcomed, and could be one of the activities that the created charity does. So far, the promise has been working via word of mouth - if you need help, you ask Promisers that you know, and they ask Promisers that they know.
Cool about the notion page. Maybe you could link it?
Great! What questions do you have? What makes you feel skeptical?
Would anyone be up for reading and responding to this article? I find myself agreeing with a lot of it.
"Effective altruism is a movement that excludes poor people"
Ah thanks for pointing out my mistake! And yes, I read this paragraph in the article, but still couldn't work out how they could provide such a precise range