[Added August 27, 2014: GiveWell Labs is now known as the Open Philanthropy Project.]
We’re starting a new initiative, GiveWell Labs, an arm of our research process that will be open to any giving opportunity, no matter what form and what sector.
One of the major challenges of this initiative (as mentioned in the previous post) will be remaining systematic and transparent despite the very broad mandate of GiveWell labs. It’s core to GiveWell that the thinking behind our recommendations
- Comes from reasoning and principles that are applied consistently, not from whims.
- Is transparent, i.e., interested people can read up on why we made the decisions we did and judge our thinking for themselves, with as little need as possible to have trust in us.
This post lays out our plan for accomplishing these.
- Sourcing general ideas. We plan to cast a wide net initially, looking pretty much anywhere we can for general funding ideas – and/or organizations – that might be promising. Key sources will include:
- Our existing research, particularly our list of the health interventions with outstanding track records and cost-effectiveness.
- Conversations with anyone who might be a particularly good source of these kinds of ideas – particularly officers at major foundations and academics who do work related to philanthropy – as well as the referrals we get from these people.
- Going over the work of anyone else who seems to similarly have a goal of identifying top opportunities for doing good. Examples we’ve found so far: Copenhagen Consensus, Back of the Envelope Guide, and the winners of sector-agnostic competitions for limited funds (Echoing Green, Ashoka, Skoll, etc.)
- Going from general ideas to specific proposals. We will maintain a ranked list of the most promising ideas from step 1, and for high-ranked ideas, we will attempt to find the people and/or organizations who can make (and, potentially, execute on) specific proposals. At this stage we’ll just be looking, in each proposal, for rough ideas of (a) costs; (b) what people/organizations will do the execution; (c) what the basic plan is.
- Detailed investigation of proposals. We will maintain a ranked list of the most promising proposals from step 2, and for the most promising ones, we will conduct in-depth investigations similar to those we’ve always conducted for promising charities. These will include in-depth conversations with the relevant people/organizations; conversations with others in their space, particularly those who have funded them or chosen not to fund them; site visits when applicable; and requesting technical reports, budgets, and other materials when applicable.
- Recommending and funding proposals. We will attempt to get any outstanding submissions from step 3 funded. We will have $1 million to spend at our discretion if we can raise no other funds, but we expect to be able to raise more if we succeed in finding great giving opportunities.
At this point we are most interested in funding others’ ideas, and have a preference for cases where the implementing organization is the same as the organization that hatched the idea and strategy. We have the impression that much philanthropy works differently, as foundation staff design their own strategies and treat grantees to some extent as “contractors” for carrying it out; this model does not currently appeal to us, but we plan on further investigating the history of philanthropy (particularly success stories) to see whether there is more promise in this approach than we’d guess.
However, a core value of ours is that interested parties – no matter who they are – ought to be able to understand as much as possible of (a) which options we considered; (b) why we chose the ones we chose. To this end, we plan on publishing:
- Extensive discussion of the values and beliefs that are relevant to which sorts of sources we use and which areas we focus on investigating. This discussion will take place via future blog posts. We hope that anyone who reads these posts will understand why we look at the areas and sources we do, and if we aren’t accomplishing this we hope our readers will comment to let us know.
- A list of the sources we use to generate ideas (step 1), along with detailed notes from particularly noteworthy conversations. Our goal here is to cast the net wide, so if you know of promising sources of ideas that fit with the values/principles we’re expressing and you don’t see us using them, we encourage you to comment.
- Discussion of the general beliefs (and relevant facts) that lead us to discard certain ideas from step #1 while moving forward to step #2 on other ideas, again via the blog.
- A full list of the proposals we consider (step #2), along with notes from discussing these proposals.
- Discussion of the general beliefs (and relevant facts) that guide our choice of particular proposals (step #2) to move to the “deep investigation” phase (step #2), again via the blog.
- Full details of the materials we collect via deep investigation (step #3) and our notes on the strengths and weaknesses of each giving option that makes it to this stage.
We will withhold information when necessary to respect confidentiality agreements. However, we will make our best effort to obtain clearance for – and share – all important/relevant information. This is the same policy we’ve used in charity investigations, and while some information remains confidential, we’ve still published the vast majority of the information we have (enough so that our views generally don’t need to be taken on trust).
- A focus on finding the best giving opportunities in terms of positive impact, rather than in terms of telling compelling stories or making donors feel good.
- Recommendations that are transparent enough to allow outsiders to draw their own conclusions and give meaningful feedback.
If we can preserve these things while working in a more open-ended way, we’ll be able to find better giving opportunities and to demonstrate our principles’ broad applicability. This means there will be fewer reasons than ever for other large givers to be keeping their own processes opaque.