insights from the shower: MDL ; proving good intentions in tutoring contexts

[gusl]

I've just had a shower. This was a good shower, as I had two important insights:

* the idea of "learning as compression" is closely related to the methodological principle that says "choose the smallest set of assumptions that explain the greatest set of observations" (I have no good name for this principle).

* when I'm learning a programming language, and I ask people questions, my intentions are frequently misinterpreted. This happened a lot on the #lisp chat channel: I would make up some example code and ask how it could be made to work, how it could be improved. People frequently replied by saying that they won't do my work for me, as if I were using their programming labour for free. I always found this perplexing, since the best way to learn programming skills is by demonstration, and the best way to understand concepts, and where they fit is interactive discussion. ¹ My intention is usually to take examples and generalize. Finding the right example in books isn't easy today.

I have also observed that the more work I put into describing the context and phrasing my question precisely, the more likely the answer will be something like "I don't answer homework problems".

Could this be a tragedy of the commons? From their point of view, I suppose it is more a question of me proving that I am putting in effort, and not just using them.

I wish people engaged in collaborative tutoring more often. Is there such a thing as a "learning cooperative", in which one could keep free riders away, through a reputation system?



^{1- Besides, one problem with telling someone to read something is that you're no longer in the same page: the tutor has no model of what the learner got out of the reading.}

Comments

[kvschwartz.livejournal.com]

But there ARE a fixed set of observations.

Anyway ... would you be happier with "analog"?

[gustavolacerda.livejournal.com]

I would be happier with "refinement".