Post-Mortem: ACX Forecasting Impact Mini-Grants
This blog raised some money! + Other people have cool project ideas too.
Scott Alexander has previously written on his blog Astral Codex Ten (ACX) about fundraising for charitable projects with a novel mechanism called impact certificates, which are sold on impact markets. To test this mechanism, and hopefully raise funds for some useful projects, the creators behind Manifold Markets built Manifund, a marketplace for this kind of fundraising, and Scott committed a certain amount of funds to be used in purchasing impact certificates at a later date which incentivizes initial funding. The about page of Manifund has the implementation details, and I feel like I only 90% understand them, so I’ll leave explaining how everything works as outside of scope for this article.
I wrote up the continuation of this blog mainly as a cheap way to contribute to the experiment and possibly get feedback on my planned work here. Surprisingly, investors valued this blog generating an average of 1 post a week for the next 6 months at >$1k of charitable impact! Raising $1k in funds definitely helps me to prioritize this blog over other projects, and hopefully being able to point to this result helps establish trust with potential collaborators in the future. I’m even more excited about the impact of this blog being evaluated in ~6 months and somehow being quantified!
Applying for funding also helped me to realize that I should be cross-posting to the EA Forum so that people who would find my work useful and/or people who would provide feedback on it are more likely to find it. I’m not sure yet exactly which posts I’ll be doing this for, but “Research” and “Application” post categories seem like an obvious fit. I started with my research summary on prediction polling, and I’ll steadily go through my backlog of posts until I’m caught up and then posting roughly concurrently for whatever topics I feel that’s useful for.
The Manifund experiment seemed very successful in surfacing some awesome forecasting related project ideas! The rest of this post will highlight and link to the three projects that I’m most excited about (AKA, they’re relevant to this blog’s mission), but follow the links for the full project details. I’ll keep tabs on how these play out, and report back in ~9 months at how well this fundraising method seemed to work for not just surfacing projects but getting them done.
Subsidize Real Money Prediction Markets on High Impact Topics - Ezra Brodey
This project would use invested funds as a cash prize pool for real money prediction markets. Subsidizing the markets in this way should mitigate to some extent the weakness of prediction markets over long time horizon forecasts that I’ve discussed here.
This project strikes me as a sort of rough proof of concept for the ability to pool funds into publicly beneficial forecasting prize pools, which made me think that pure crowdsourcing might be even better!
Manifold feature to improve non-resolving popularity markets - Kenan Schaefkoker
An experiment in using visual status badges as rewards for trading in otherwise never going to resolve play money markets that are very long term / infinite / unresolvable. The developer also hypothesizes that this could foster conversation with these forecasters on those topics, allowing them to outwardly signal their interest.
This seems like an inexpensive experiment in trying to mitigate one of the downsides of prediction markets. I believe it’s only possible because Manifold is open source. Very cool!
Comparing forecasting platform accuracy - Jack
An experiment in comparing the accuracy of different forecasting platforms, understanding differences in accuracy, and even aggregating forecasts across platforms to improve accuracy. This is just obviously awesome.