Walking the knowledge graph

All timestamps are based on your local time of:

Posted by: stak
Posted on: 2011-11-12 01:45:09

Over the last couple of years I've noticed an interesting phenomenon - my list of books to read seems to be growing almost exponentially. I find that every book I read leads to me adding a few more books to my list because they branch off into interesting related areas from what I just read.

In my head I visualize this as a graph where you start at some node (i.e. a subject area) and traverse outwards to connected nodes (other subject areas). What's interesting is that in the frontier of nodes that you have not yet visited (i.e. the subject areas most related to what you already know), the likelihood of you visiting a node (reading up on a subject) increases with the edges that connect it to your already visited nodes (how much it interacts with things you already know).

The main consequence of this, and what is resulting in my exploding reading list, is this: the more you read and learn, the more interesting new things become. You end up being able to relate the new subject areas to things you already know, and the more you know, the more this happens.

Another thing that's interesting about this process is that if you start at a subject area that you're passionate about, you have an intrinsic motivation to learn about related subject areas. This means you visit frontier nodes more efficiently, and reach the point of "everything is interesting" sooner. (This is defined to be when your reading list consists of every book you haven't already read.)

In contrast, if you are forced into the graph at a particular node (say by taking university courses you have no interest in), then you have no motivation to reach out to other adjacent nodes. This effectively kills your desire to learn, unless you are able to start over from another node that you do have an interest in. Unfortunately that seems to happen pretty rarely.

There are of course scenarios in between these two extremes as well - it may be that the area of the knowledge graph you're in is fairly sparse and so your visited nodes don't connect to a lot of other nodes. In this case you might have a harder time finding related areas that interest you. In that case you could just pick the thing that seems most interesting, or even just pick randomly, and hope for the best. As you visit more nodes, you're bound to hit a denser part of the graph sooner or later.

Posted by varun at 2011-11-12 23:10:03
The William Gibson summary of your problem (and mine) is - "all things are increasingly interesting to me".

Btw, unrelated - what stops you from using recaptcha as opposed to the custom captcha? Not that 52+43 is hard, but curious why.
[ Reply to this ]
Posted by stak at 2011-11-13 10:30:10
I don't like recaptcha because (1) it requires fetching content from an external server, which increases injection attack surface area, (2) it doesn't work with javascript disabled (at not last time I checked, which was a long time ago), and (3) recaptcha is owned by Google, and I'm trying to reduce dependency on Google rather than increase it.

FWIW, all of my captchas are designed to have no carry in the addition, so they should be pretty easy.
[ Reply to this ]

[ Add a new comment ]

(c) Kartikaya Gupta, 2004-2024. User comments owned by their respective posters. All rights reserved.
You are accessing this website via IPv4. Consider upgrading to IPv6!