Back in 2017, I started doing something I called Epistemic Spot Checks, in which I would read the first few chapters of a book and investigate their claims (factual or logical), to see how much they could be trusted. I now think this was a suboptimal way to do things, but it’s how the ball got rolling. You can read these investigations in the epistemicspotcheck tag.
This eventually grew into the Knowledge Bootstrapping Project, which has the goal of turning questions into answers without undo deference to credentialed authority (posts on this still fall under the epistemicspotcheck tag for historical reasons, or metaESC if they are about the process itself). This culminated in an algorithm for investigating questions as efficiently and thoroughly as possible, v0.1 of which can be found here. Some external reviews of this process can be found here (with follow up) and here. This work was originally supported by a grant from the Long Term Future Fund.
But it turned out that the key obstacle to learning for me was not an algorithm, it was emotional blocks around the concept of learning and forming my own opinions, and resolving those is a process all its own. My current project is studying that process, although due to the nature of things that still involves a lot of object level research.
My goal is to teach both the object level and the emotional process to others, and in doing so refine it to a more generalizable method.