I think the first point is mostly out of the scope of a personal kb. But it fascinates me.
In the ideal world, I could see this happening in the following phases:
- I as the user, specify categories and subcategories that i am interested in, the personal KB system can access some external API and provide me with new readings. In this case it is similar to a recommendation engine. I can like/dislike, this can alter the future recommendations.
- The system based on my learning style provides me with recommendations (no user specified categories). Assume some external API as well.
- The system based on some globally available knowledge map design NEW map to learn for me. This is no longer individual kb, but a fully linked kb map. The system guide me through it.
2, 3 are at least years away from us (and obviously not part of a personl kb system)… but #1 should be easy if such an API exists. Again I think personal kb system should focus on organizing and visualizing kb points, #1 is just nice to have, not a essential function for a personal kb.
Now i think back about pruning stale kb, i think it is tricky to do since everything is linked. Deleting one could result in many disconnections. Human memory mechanism I believe (I am not certain) uses association very efficiently (associative memory?). So keep stale nodes maybe more beneficial than removing it. If n nodes are connected with 1, it roughy takes n+1 in memory, but if we remove the 1, then it might be n² in the memory?