You are asking, gwern, a question about what people like about your writings. My answer, I hope, will be taken as a friendly suggestion about how to make your writings even better and more likeable. One area of research interest you and I share is the nature of human intelligence and what might be possible to do to improve human intelligence at the individual level. Having read extensively about this issue since 1993 (arguably since 1989, when I began extensive research on early childhood education), I been collecting bibliographies of the best sources on this topic for a long time.

When I started editing Wikipedia as a registered Wikipedian a few years ago, I soon discovered that many articles there on related topics are compiled in complete ignorance of the best sources on those topics. To do something about that, I have been compiling a bibliography in Wikipedia user space on IQ and human intelligence,

http://en.wikipedia.org/wiki/User:WeijiBaikeBianji/Intellige...

and I encourage you to look there for recommendations of up-to-date reliable secondary sources. (I am updating that bibliography as I do more research, so there may be new entries added there fairly soon.) It would be interesting to see what new paths your writings will take after more digestion of the previous literature on our shared topic of interest.

Another friendly comment is that many of your readers are glad to see your statistical approach to some of the issues you write about. I too like well writings that apply statistical analysis to data once they are gathered. I like even better writings that apply statistics to examine whether data are adequate to the task of answering the question posed by a researcher, as statistics is the science of data,

http://statland.org/MyPapers/MAAFIXED.PDF

and what I have found out by participating in the University of Minnesota's journal club in behavior genetics with leading researchers on that subject is that many findings on human behavior now need to be reexamined as statistically astute psychology researchers reexamine the quality of data in old studies. Being aware of issues of validity of inference

http://www.escholarship.org/uc/item/6hb3k0nz

and how researchers can fool themselves

http://calteches.library.caltech.edu/51/2/CargoCult.pdf

is even more important than applying statistical manipulations to a data set after the data set is gathered. In their best use, statistics can help show which data sets need to be reexamined to make sure that a study inference is really warranted by data.

http://www.p-curve.com/