Manjari Narayan

So you have a principal-agent problem. You need institutional management that is somehow accountable to you to spend your money well. However, you know you don't have the expertise to assess that, at least not on a timescale less than 5-10 years.

The management has two functions: straight-up operations (make sure your building's chalkboards are supplied with chalk and the employee health plan is sound), and research management per se, which includes line management of scientists (if you do that) and hiring or grant-making

This is an inherently meta-rational function, meaning that it's about understanding and acting upon rational (scientific) work from above and outside it.

Meta-rational also meaning that it inherently can't be done rationally. For example, it cannot be done by metrics of any sort (although it might take metrics, probably peculiar and continually adjusted ones, into account). It's inherently a matter of judgement, not justification.

This means that decisions about what research to fund must NOT be accountable to anyone other than, in the long run, the funder(s), and must NOT be justified in any specific way. That would be merely-rational and leads to the bureaucracy-heavy mediocrity we have now.

The danger here is that the decisions are based on cronyism, self-interest, or (more likely) the hobbyhorsical crackpot whims of some charismatic charlatan. (Not pointing to any specific examples here :)

Like all principal-agents problems this is inherently not solvable by any specific mechanism (it is also meta-rational). However, the good news is SV VCs make meta-rational decisions, and character decisions, all day everyday, and may be pretty good at them.