Key writings and resources

Sept-Nov. 2022: GiveWell 'change our mind' contest

Perhaps motivated by critiques, suggestions, and further analysis like those in this gitbook and the writings below... In Sept. 2022 GiveWell announced a 'change our mind' contest, with $35k in total prize.

A great deal of additional writing and work has been posted in response to this, much of which engages the issues discussed here, including Incorporating uncertainty, transparent and organized data and calculation 'pipelines', and user input of moral and other parameters. We plan to outline this work below (to do).

Tanae Rao's work, focusing on AMF

Squiggle notebook:

Adding Quantified Uncertainty to GiveWell's Cost Effectiveness Analysis of the Against Malaria FoundationObservable

E

EA Forum brief on this here

Froolow's discussion & ex. 'refactoring'

Discussion of the approaches to uncertainty and GiveWell's processes

Methods for improving uncertainty analysis in EA cost-effectiveness models - EA Forum

Refactored GiveWell model v9.xlsmDropbox

Excel-based model, multiple charities

See also Froolow's more general discussion/critique of GiveWell's processes here

Sam Nolan's (Quri) work, focusing on GiveDirectly

GiveWell's GiveDirectly Cost Effectiveness Analysis

First-pass implementation on 'observable'

Quantifying Uncertainty in GiveWell's GiveDirectly Cost-Effectiveness Analysis - EA Forum

Cole_haus (earlier) work

Uncertainty and sensitivity analyses of GiveWell's cost-effectiveness analyses ... by cole_haus using Python code (full model on Github here

Uncertainty and sensitivity analyses of GiveWell's cost-effectiveness analyses - EA Forum

Adjacent: Pedant/Hazelfire

Pedant, a type checker for Cost Effectiveness Analysis - EA Forum

Further Background and explanations

I think this 'explainer' is a step in the right direction in some ways.

Why Is It So Expensive to Save Lives?GiveWell

The general case

How to Measure Anything: Finding the Value of Intangibles in Business

Douglas Hubbard, strong 'business/layman' arguments for explicitly stating modeling uncertainty and better calibrating one's own beliefs