Comments
As you might know, we’ve recently launched the Rethink Priorities Cross-Cause Fund (CCF), a pooled fund that allocates donations across high-impact giving opportunities in Global Health and Development, Animal Welfare, and Global Catastrophic Risks. Thanks to generous donors and supporters, we have now completed our first distribution round, allocating more than $700,000 across the funds in our model, following its current recommended split.
The CCF exists because cross-cause prioritization - comparing giving opportunities across different causes on a single scale - remains underdeveloped in effective giving. Rather than dividing funding by cause area first and optimizing within each one separately, the CCF uses an explicit model to evaluate specific funds directly against one another, so that a marginal dollar goes where it is likely to do the most good, all things considered, regardless of cause area. For the full rationale behind the fund, please read our launch announcement.
In the first round, we distributed more than $700,000 across the funds included in our model, following the current recommended allocation that you can follow on our website. The table below shows how this round's distribution broke down by fund:
We are grateful to the donors who trusted our model and whose contributions to the CCF made this first distribution possible.
We selected funds for the CCF using five criteria: coverage across our three cause areas, confidence in a fund’s quality and track record, whether we could obtain or model reliable cost-effectiveness data, relevance to the funding decisions we expected our early users to face (a consideration that will carry less weight in future versions), and legal restrictions, since as a 501(c)(3) public charity, RP cannot direct funding toward political candidates or parties.
The model doesn’t rely on one fixed set of moral or empirical assumptions. Instead, it draws on a range of worldviews, each combining a set of ethical commitments with beliefs about how the world works. Every worldview is also assigned a credence, which determines how much weight it gets in the final result.
Every worldview needs a common way to compare different kinds of impact, such as a life saved, proportional income gain, or reduced animal suffering. To do this, the model measures everything against a shared baseline of one human life-year and converts other outcomes, including disability-averted life years, proportional income gains, and reduced suffering across species such as chickens, fish, and shrimp, into that same unit.
Worldviews also differ in how much they value impact that occurs far in the future compared to now, and in how much they favor safer, more predictable outcomes over riskier ones with a higher potential payoff. Each worldview sets its own position on both. Worldviews that assign meaningful probability to a near-term AI catastrophe also apply a discount to funds that don’t directly address AI risk, reflecting the chance that such an event could reduce the impact of other interventions.
To generate a recommendation, we score every fund under each worldview using our fund-by-fund cost-effectiveness estimates, then combine those scores into a single portfolio using different aggregation methods, since no single way of combining differing moral views is universally accepted as correct. The model also accounts for diminishing returns: a fund’s most cost-effective opportunities tend to get funded first, so each additional dollar a fund receives tends to have somewhat less impact than the one before it.
For the full set of default inputs, our detailed fund-by-fund cost-effectiveness data, and a walkthrough of these steps in action, visit the CCF methodology page.
This is our first distribution, and we want to be upfront that both our fund coverage and our underlying model remain works in progress. We plan to keep refining our methodology and cost-effectiveness estimates, and expand the range of funds and cause areas we evaluate over time.
If you would like to see how your own moral and empirical assumptions shape a cross-cause allocation, we encourage you to try Donor Compass, an interactive tool that runs your own worldview through the same model, producing a personalized allocation across the funds above.
And if you have feedback on this round, this update, or the model generally, we would like to hear from you. You can contact us at: [email protected].
If you are interested in contributing to future rounds, you can donate to the CCF either as a one-time donation or on a recurring basis via our website. Your contribution is pooled with other donors’ and allocated according to our current recommended split, the same process described above.
ContributeWe are grateful to everyone who chooses to support this work.
Conflicts of interest statement
Rethink Priorities (RP) has received grants from Longview Philanthropy, the EA Animal Welfare Fund, the Navigation Fund, and Coefficient Giving, and has done commissioned research for GiveWell and Coefficient Giving.
RP plays no direct role in these organizations’ grantmaking decisions and conducts independent cost-effectiveness analyses of every fund included in the CCF.
