The rules involve integrating a lot of fairly abstract tendencies and applying them to human behavior. A person who tends to idealize male figures and likes to learn by via physical participation and who has a fairly short attention span might be the derived interpretations from the first even more abstract level. Then you need to integrate these into giving a description of how this person would be in public school, say. I think that kind of thing would be very hard for computers to do now.
Now computers could start doing enormous statistical analyses, and learn something over time. But I think a vastly better approach would be to do exactly as they first did with chess computers which is to use strong players to create heuristics so that the number crunching advantages can really play out. Even now combined human/computer teams beat the best computers. Someday computers will be able to crunch the whole damn thing of chess, but they aren’t there yet.
In related news, it appears that your study design has actually been done:
Actually what I read there is the the CPI test may be poor. Please read that abstract again.
We would not have concluded that they had predictive power if the single study had shown positive results, that is something we can be sure of. And while Nature, is, yes a very respectable journal, it is also much more likely to be biased towards things that fit their own paradigms, and to be less critical of papers that fit than those that don’t. To me that abstract screams not remotely conclusive of anything due to the problem of the CPI potentially being the problem.
Thank you for bringing that article to my attention. I have been surprised that such a testing protocol hadn’t been tried. I have also tried to interest both astrology organizations and scientists in precisely that protocol and no one has been interested. I do understand some of the motives on both sides, but sometimes it seems people would rather babble without much knowledge of the other side for decades and decades then to try to actually resolve things.
It seems silly that they did not retest using some other personality description than the CPI, one that people recognized better than chance. I mean, if you are interested in drawing a clear conclusion. Imagine a medical paper trying to find out about a treatment for a disease and using as a core part of the experiement a test for the disease that might not work at all. That would be laughable.