Hybrid Forecasting of GeoPolitical Events

in #politics3 years ago (edited)

This is a project by IARPA (like DARPA but for the intelligence community) where they are attempting to combine crowdsourcing with machine learning.


I signed up and took the test, which was in three sections:

  • True/false statements about geopolitics, where (I'm guessing) the false answers were generated by randomly replacing one of the nouns with something plausible.  Then I had to provide my confidence level.  This is actually really similar to how psychology researchers would design experiments, and now it's spread to algorithmically generated homework for companies like Cengage.
  • Visual pattern recognition, where the patterns were line drawings that progressed in color, rotation, or some other property.  Psychologists call the simple version of this "seriation."

 [image link to Wikipedia]

  • Numerical pattern recognition, with sequences of numbers following some unknown rule.  I was less confident about this one, but hey, it's a test.  I've taken lots of tests, and it's not like they're going to come and get me if I fail, right?

 [image link to Wikipedia]


The directions said I would get some feedback, but I haven't seen it yet.  Maybe it will come by e-mail, in which case I'll post it in the comments.  Or maybe it will be in the form of a personal visit from my paranoid friend up there, in which case I won't remember anything.  But the blockchain will . . . unless there's an alien equivalent of the neuralyzer for computers.

In any case, the next step is some training, which will supposedly raise my ability to forecast geopolitical events.  I'll report on that next time.  I think this is going to be fun.