Another fun shower thought I had earlier today, was about the problem where researchers (read: me) could be disincentivised to work on or publish something because, once it's published, prospective funders aren't going to pay for it. As usual, they'll reason that their purchase has no counterfactual impact if something is already done.
Solution: Do your work ahead of time and hold it in hostage!
Ideally, I could just go up to a funder, tell them "hand me X money and I'll publish or let you read this work I did." Now, I can *demonstrate* the work before they pay for it, meaning trust and credentials has a lesser role to play. Unfortunately, if I show them my work before they pay for it, it obviates their incentive to pay. So I'm back to having to accept their reduced offer which will be heavily penalised by uncertainty about my competence.
So instead, both me and the potential funder(s) should pay a mutually-trusted third-party to valuate the work based on the funders' subjective values. They're committed to respecting IP.
Yes! Another reason to worry about narrow targets is costs of compromise*, and researchers are often better equipped to judge their own abilities, interests, and which questions are highest-impact for them to research. Rarely will they find a prize that corresponds to what they can most effectively contribute with.
Re Polis: It may help against the defect-defect coordination problem you talk about, but it doesn't solve the cost-of-compromise problem. The most upvoted research questions are still very narrow from the researcher's perspective.
Another fun shower thought I had earlier today, was about the problem where researchers (read: me) could be disincentivised to work on or publish something because, once it's published, prospective funders aren't going to pay for it. As usual, they'll reason that their purchase has no counterfactual impact if something is already done.
Solution: Do your work ahead of time and hold it in hostage!
Ideally, I could just go up to a funder, tell them "hand me X money and I'll publish or let you read this work I did." Now, I can *demonstrate* the work before they pay for it, meaning trust and credentials has a lesser role to play. Unfortunately, if I show them my work before they pay for it, it obviates their incentive to pay. So I'm back to having to accept their reduced offer which will be heavily penalised by uncertainty about my competence.
So instead, both me and the potential funder(s) should pay a mutually-trusted third-party to valuate the work based on the funders' subjective values. They're committed to respecting IP.
Meh, it's dumb and I should go to sleep.
Yes! Another reason to worry about narrow targets is costs of compromise*, and researchers are often better equipped to judge their own abilities, interests, and which questions are highest-impact for them to research. Rarely will they find a prize that corresponds to what they can most effectively contribute with.
Re Polis: It may help against the defect-defect coordination problem you talk about, but it doesn't solve the cost-of-compromise problem. The most upvoted research questions are still very narrow from the researcher's perspective.
*https://www.lesswrong.com/posts/DdDt5NXkfuxAnAvGJ/changing-the-world-through-slack-and-hobbies