I think this is obfuscating the good points, I appreciate many of the points but they seem to be ticked off rather than front and center.
I am afraid the frame of "When to" is promoting a binary mindset which is fundamentally opposed to proper decision making.
I am reading it as attempting to have decision points for when to collapse distributions to point estimates. "Use of explicit probabilities"
You always have the explicit distribution. You always have a policy (why didn't it say policy in the alliterative p title) You always break apart the timeline and draw the causal dag.
This is offensive to reasonable planning: "Some creatures would be better served by mapping out the dynamic dependencies of the world" Always draw the dependencies!
The question is when to communicate the point estimate versus distribution. When to communicate the dependencies or just the final distribution.
People allege the crazy train when you are imagining a point estimate represents the best information that is used for making a decision. That is the implicit suggestion when you discuss point estimates.
Quick suggestions, communicating a point estimate is poor:
Poker is unnecessary for the analogy, just probability of a draw from an urn.
We are speculating on how many balls are in the urn when a much better question would be Given we get the urn will we know how many balls are in it? How much does that cost? Can we do things before opening the urn that change the contents? How much does that cost?
Can we not sign for the urn when the Amazon delivery guy arrives? How much does that cost?
Ok that is a joke, but the idea is that we don't know what recourse we have and those actually are important and affect the point estimate.
The probability is downstream from certain decisions, we need to identify those decisions that affect the probability.
Does that mean the point estimate is useless, well maybe because those decisions might reduce the probability by some relative amount, ie if we get congress to pass the bill the odds are half no matter what they were before.
If you go, yeah but I say it is 27.45% and 13.725% is too high. They a decision maker goes "Sure, but I still want to halve it, give me something else that halves it stop telling me a number with no use"
You mention relative likelihood, but it is buried in a sentence of jargon I just had to search for it to remember if you said it.
Finally, frame the analysis relative to a better considered approach to Robust Decision Making, a la Decision Making Under Deep Uncertainty, not relative to Scott or Holden's world view which are just poor starting points.
"While useful, even models that produced a perfect probability density function for precisely selected outcomes would not prove sufficient to answer such questions. Nor are they necessary."
I recommend reading DMDU since it goes into much more detail than I can do justice.
Yet, I believe you are focusing heavily on the concept of the distribution existing while the claim should be restated.
Deep uncertainty implies that the range of reasonable distributions allows so many reasonable decisions that attempting to "agree on assumptions then act" is a poor frame. Instead, you want to explore all reasonable distributions then "agree on decisions".
If you are in a state where reasonable people are producing meaningfully different decisions, ie different sign from your convention above, based on the distribution and weighting terms. Then it becomes more useful to focus on the timeline and tradeoffs rather than the current understanding of the distribution:
Explore the largest range of scenarios (in the 1/n case each time you add another plausible scenario it changes all scenario weights)
Understand the sequence of actions/information released
Identify actions that won't change with new info
Identify information that will meaningfully change your decision
Identify actions that should follow given the new information
Quantify tradeoffs forced with decisions
This results is building an adapting policy pathway rather than making a decision or even choosing a model framework.
Value is derived from expanding the suite of policies, scenarios and objectives or illustrating the tradeoffs between objectives and how to minimize those tradeoffs via sequencing.
This is in contrast to emphasizing the optimal distribution (or worse, point estimate) conditional on all available data. Since that distribution is still subject to change in time and evaluated under different weights by different stakeholders.
First, at Malengo the students fully fund the next cohort via repaying the original donation in an ISA.
This means that funding 1 student will actually fund many students over time. Using the numbers above you get a rate of return around 6% annualized. So funding a student is sorta infinite students 0% discount rates. But that is unreasonable, so let's just cap at the next 100 years and say 2% discount rate from inflation.
BOTEC for 1 funding pays 12.5 students or a student every 8 years.
That changes your calculation from 3x givedirectly to 37.5x.
Second, you also said the students are richer but that is factually incorrect, the program is means testing to ensure that students are well targeted.
Finally, there are other fudge factors, but they are all dwarfed by the development benefits of immigration.
https://www.nber.org/papers/w29862
This shows that nearly 80% of long-run income gains are accrued within sending countries across a wide variety of channels.
Hence, I think 37.5x GiveDirectly is a completely reasonable estimate.
https://twitter.com/jhaushofer/status/1685541903813804033?s=20
"Malengo sent two pilot cohorts of students from Uganda to Germany (6 in Fall 2021, 17 in Fall 2022). All of these students are currently in Germany and making progress towards their degrees.
"