Comment by arjunnarayan | Hacker News Reader

Comment by arjunnarayan | Hacker News Reader

Parent Full thread

I'm not surprised. The "logic" section of the LSAT (traditionally said to be the hardest part of the test) is much easier if you know some computer science/discrete math. A lot of questions can be quickly parsed into some sort of constraint satisfaction problem, and using the right tool (drawing truth tables/constructing a directed acyclic graph) means you can get to the answer much faster.[1] Since the LSAT is a time-pressure based test, this gives you a big advantage. Of course, the reading comprehension parts are still hard, and engineering gives you no advantage there. There's also lots of selection bias potentially going on - engineers who go on to take the LSAT are rare. Who knows how that prior correlates with ability.

Somewhat relatedly, it's been well known that for CS PhD admissions, at least with native English speakers, the GRE Verbal is a better predictor of PhD performance than the GRE Math. It's unclear how much of this is top-end compression in math though (almost everyone going to a top CS PhD school gets a 790 or 800 on the GRE Math, so there's no signal left there, whereas verbal has a higher spread. A 720 on the verbal is 98th percentile).[2]

[1] I know this because in 2009 as a college senior, I took some practice LSAT tests, and found the logic section endlessly delightful and challenging in a timed setting - I was this close to applying to law schools before I changed my mind and did a CS PhD instead. I know it sounds like a weird choice, but 2009 was a weird soul-searching year for me for various personal reasons.

[2] All of this has changed since 2011 when the GRE scoring system changed to a 130-170 scale from the old 200-800 scale, along with changes to the test itself. All of my points are probably obsolete.

I think the section that delighted you is actually called the "analytical reasoning" section. The "logical reasoning" section gives you a paragraph about some subject, and that asks about inferences and deductions you can make from that.

There are samples from both sections here for comparison: http://www.lsac.org/jd/lsat/about-the-lsat/

I too found analytic reasoning delightful, both during practice tests and during the actual test. One thing about the actual test that kicked it up a notch over practice tests (aside from the pressure of it being for real) was that you could not bring scratch paper to the real test. There was a fair bit of empty space in the test booklet so you had room for truth tables, graphs, and such, but if you botched one and had to start over you might find scratch space tight.

I aced analytic reasoning. Shortly after the test, I saw a magazine at the supermarket full of those kinds of problems. The shelf position and the ads in the magazine indicated it was aimed at older women. I bought it, because I enjoyed that part of the test so much, but didn't have much hope because I aced that part of the LSAT! How are puzzles for Grannies going to be a challenge? So I tried a level 3 difficulty puzzle (range was 1 to 5 with 5 the hardest).

It DESTROYED me. I could do level 1 reasonably, and level 2 with a lot of effort.

I vowed that if I did become a lawyer, I would not take cases that would pit me against old ladies. (I did not become a lawyer. I decided near the end of law school that I would rather be a programmer who knew a lot about law than a lawyer who knew a lot about programming. My policy of not going against old ladies in battles of wits remains in force).

The GRE math is actually a dumbed down version of the SAT math so it's not surprising there's no signal left among engineering students. Probably most engineering students have gotten better at math since high school.

The GRE verbal is a great predictor of success in CS grad school and any kind of science grad school because it's so heavily g-loaded. Verbal scores are very hard to game and correlate well with raw IQ even when people try to study for the test and artificially improve scores. Antonyms, vocabulary, and reading comprehension are very good measures of the same mental ability that deep cutting edge research requires.

Of course, that doesn't work for non-fluent English speakers.

Interesting to read that. I have always thought of CS as a discipline that involves creating and testing competing explanatory models of the world similarly to the way that a lawyer might construct an argument to represent a series of events or a mathematician might construct an equation to represent a set/series of data.

You could also think of law as a form of code.

I think it's more like an api. You hack it to get what you want.

CS is also about creatively using fixed rules to get a desired result.

But then, so's engineering.

Standardized test scores aren't that important after your first professional job. For some reason, though, the financial industry is infatuated with GPAs.

They get an unspeakable number of applicants and finance requires horse like work ethic, a high tolerance for boredom and detail orientation. A high GPA is a strong signal of all of those and also has a very high correlation with g/IQ/processing speed.

Work sample tests might work better but at least in the UK they do multiple rounds of structured interviews followed by free form interviews after filtering by class of degree/university prestige. I bet their false positive rate on hiring is very low even if the false negative rate is high.

To your [1], I actually think studying for the LSAT has been the most intellectually enjoyable part of my legal career so far!

I've heard the same thing from Philosophy majors though...

Reply on news.ycombinator.com