[Link] Research as a Stochastic Decision Process


Via Gw­ern.

In this post I will talk about an ap­proach to re­search (and other pro­jects that in­volve high un­cer­tainty) that has sub­stan­tially im­proved my pro­duc­tivity. Be­fore im­ple­ment­ing this ap­proach, I made lit­tle re­search progress for over a year; af­ter­wards, I com­pleted one pro­ject ev­ery four months on av­er­age. Other changes also con­tributed, but I ex­pect the ideas here to at least dou­ble your pro­duc­tivity if you aren’t already em­ploy­ing a similar pro­cess.

Many EA type ac­tivi­ties could benefit from this frame­work!