Posts

Comments

Comment by Frederic Bahnson (fbahnson@gmail.com) on CFAR Takeaways: Andrew Critch · 2024-02-23T22:43:42.727Z · LW · GW

The filtering described here seems moderately specific but not sensitive, whether or not you agree with the "income implies competence" relationship being strong.

It seems true that those who are interested in and can pay for a $4k course of this type are more likely to have 17 of the attributes in question than a person picked at random from the population. However, the filter tells you nothing about, and completely excludes, a large number of people who would fit the "have 17 of these attributes" criteria but not the "have $4k to spend on a course or the time to take it" criteria. 

The filter allows in a population of people with above-average chances of meeting the attribute criteria, but blocks a large and unknown number of other people who would also meet that criteria. 

It is potentially good for creating a desired environment in the course (having mostly people with a lot of the desired attributes), but is not a good filter for identifying the much larger population of people who might be interested in and benefitted by the course (as described in article as having 17 of the attributes and therefore capable of picking up the other two).