Hey Sophie! Thanks again for the nudge to post here.

You're right that the most movement-driven work resists clean counterfactuals. There's no control group for Extinction Rebellion, and forcing FundAction's grantees into a sample-size study could distort the thing and likely annoy the groups themselves. An RCT doesn't fix that.

But I'd push us on the binary. The choice isn't statistics or "feelings dreaded by all EAs." It looks that way because the field keeps reaching for bednet-style tools, watching them fail to fit, and concluding the thing itself can't be measured. Trust can be studied rigorously. Social network analysis, mixed methods, racialized-trust frameworks.

I'm working on this right now with Dr. Amber Banks. Her PhD mapped cross-cultural trust networks using network analysis, and she spent years as a program officer inside Gates. She doesn't experience rigor and trust-based knowing as opposites.

"Fund rigorous evaluations" doesn't mean RCTs everywhere. It means stop treating proximity as the place evidence ends. Often the better question isn't "did it beat a counterfactual" but "did this process surface and fund what the legible process couldn't see?" You can actually measure that.

And you're right that underneath all of it sits the real question: whether EA will count relationship-based knowing as evidence at all. That discomfort your wink points at is the content, not a side issue. Amber and I are betting you can honor that knowing and still study it well. We're taking a crack at it.

One of my off-the-wall ideas for the donor intelligence platform I'm building, Kindora, is a follow button for philanthropy where we create a highly transparent way of piggy backing on the diligence of proximate funders. I'll likely build a prototype of it at some point this year.

Thank you for the thoughtful read. I resonate with the "we don't even know we don't know" point because a portfolio can look well-optimized even when the unfound orgs never enter the denominator.

The search cost isn't just high, it's front-loaded and shareable in a way the field doesn't take advantage of. Once a grantmaker has done the work of mapping who's real in a place, that map is close to a public good, but too often we fund as if every funder has to rediscover it alone. A shared diligence layer for high-proximity orgs would change the per-grant math.