gwern comments on Open Thread, June 1-15, 2012 - Less Wrong

3 Post author: OpenThreadGuy 01 June 2012 04:01AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (252)
Tags: open thread open_thread

Comments (252)

You are viewing a single comment's thread. Show more comments above.

Comment author: gwern 12 June 2012 07:54:27PM 1 point [-]

There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.

And how does one do that? The problem is that most historical facts are publicly available, so how does one distinguish a theory producing by data mining and overfitting from one that wasn't? The only historian I can think of who has anything close to an answer to that is Turchin via the usual statistics method of holding back data to test the extrapolations.

Turchin and Carrier are discussed occasionally, but not that much; why should I think this is not the right amount of discussion?

Comment author: TimS 12 June 2012 08:06:52PM 2 points [-]

The bigger problem with most historical analysis takes the following form:

1) Pick a historical thesis (usually because it supports one's pre-existing moral positions)
2) Find all historical evidence that supports that theory
3) Throw any remaining historical evidence in the trash

If you have successfully avoided that trap, congratulations. Society as a whole has not, and this community is not noticeably better than the greater societies we are draw from.

Comment author: Eugine_Nier 14 June 2012 12:33:12AM 0 points [-]

There are historical theories that actually fit most of the facts and pseudo-historical theories that fit carefully selected sets of facts. Being able to tell the difference is a valuable skill that members of this community should try to develop.

And how does one do that? The problem is that most historical facts are publicly available, so how does one distinguish a theory producing by data mining and overfitting from one that wasn't?

This is a thick problem.

Powered by Reddit Powered by Reddit