Alex N.
Comments4
This is very interesting, and has close parallels to a project I am working on. I think we share an underlying premise: effective agentic AI giving will require an infrastructure layer, not just better models.
My contribution to this space is zooidfund, a live experiment that lets AI agents discover, evaluate, and donate directly to humanitarian campaigns created by individuals in need and by organizations. Donations are direct: zooidfund does not hold or intermediate funds. It is still very early, but it is live now, with real campaigns and observable agent behavior.
I think there are important problems and opportunities here for EA: improving evidence-based allocation, bringing higher-quality decision-making to the level of individual donations, and enabling faster response and iteration than traditional funding processes often allow.
More broadly, for AI, I think this kind of infrastructure could become relevant to the question of how resources are directed as AI capabilities increase. If AI systems can help identify need, evaluate evidence, and route funding more efficiently, that could become one mechanism for distributing some of the benefits of AI more broadly, including outside existing institutional funding channels.
Would be great to connect.
A fair number of this would fall into a bucket of charity's impact on larger systemic change. Seemingly cost effective activities could result in an overall negative effect, food donations destroying local agriculture as an example. While seemingly wasteful interventions, a well organised banquet with right government official, can have an absolutely outsized second order positive effect though policy change.
These are very difficult to measure, although AI may open possibilities.
Not sure also how timing features here, a "wasteful" intervention delivered on time in a crisis can have much larger positive effect then a much better organised effort later. Delivering water to freshly displaced population in a desert, even by most ineffective methods like water tracking can have the highest ratio of life's saved per dollar donated.
I observed similar results: models do appear to have EA-adjacent moral priors. It seems debatable whether these come mainly from training data itself, from the people and institutions shaping the models, or from the way the questions are framed.
The more interesting question to me is how these priors translate into behaviour once an agent has a wallet, a budget, access to data on individual or community needs, peer signals from other donors, and a requirement to give a public reason for each allocation.
Disclosure: I run zooid.fund, a platform I built that enables AI agents to search, evaluate, and donate directly to humanitarian campaigns created by people in need. I also currently operate and fund the first agents active on the platform.
It is very early and the current scale is tiny: 14 active campaigns, 4 active agents, 30 donation events, and a total of 186.50 USDC donated so far. Some agents are more deterministic AgentKit-style setups with explicit constraints and narrow decision rules; others are more open-ended LLM/persona-driven agents. I expect these architectures to produce different allocation patterns.
I would not claim the data shows much yet. The selection effects are large, I am not independent, and donation volume is still experimental. But I think this kind of setup is a useful complement to prompt-only studies: not just “what does the model say is good?”, but “what does a scaffolded agent actually fund when it has to choose?”
The campaigns were created by real people in response to invitations in a few relevant subreddits. I am cautious about advertising the platform more broadly until donor participation increases and donation volume is high enough that campaign creators have a reasonable chance of receiving something, rather than just being asked to expose their needs to an empty market.