Hacker News Reader

Comment by WalterBright | Hacker News Reader

Caltech had timed exams (2-3 hours) and infinite time exams, at the discretion of the professor.

The students hated the infinite time ones, because nobody knew how much time other students spent on the test so one felt obliged to spend inordinate amounts of time on it.

Besides, if you couldn't solve the exam problems in 2 hours, you simply didn't know the material.

View on HN

In the upper division of my undergrad physics degree that was really common. Open book, open everything except peers. I personally loved those exams and my grades went way up. I could walk away for a few minutes if I was stuck, maybe grab a beer to relax, and get back and solve the problems. But I think this is much harder to do and getting even more difficult. I was at a small university and you really couldn't google the answers. It was really easy to write google proof questions. But a key part was that the classes were small, so it was pretty obvious if people were cheating.

I went to grad school in CS after a few years of work and when I taught I centered the classes around projects. This was more difficult in lower division classes but very effective in upper. But it is more work on the person running the class.

I don't think there's a clear solution that can be applied to all fields or all classes, but I do think it is important people rethink how to do things.

That's how my parents taught. Design questions to make the students apply their knowledge rather than regurgitate it. Forget a fact it's being applied to, look it up. Don't understand the concepts, you're stuck. Know the material, piece of cake. One time I was in my father's classroom because he was showing a film he wanted me to see. There was a quiz afterwards, he knew it wouldn't be alien to me and had me try it. 5 minutes later I turn it in, the class thinks I gave up. Then he says I aced it. But I graded an awful lot of his tests, I know that when I didn't know the material I wouldn't stand a chance. The day I found a question that I could guess was notable enough to me that I asked my mother about it. (A case of not knowing the fact. Her supplying the information that the tribe in question was a stone age culture in the New Guinea jungles made the why apparent.)

One physics exam question I remember was derive Maxwell's Equations from the starting point of presuming the existence of magnetic monopoles. This sounds like an intractable problem, but it turned out that if you really understood how they were derived, all you had to do was switch out the charge monopoles with the magnetic monopoles, and it was a piece of cake.

A similar exam problem in AMA95 was to derive the hyperbolic transforms. The trick there was to know how the Fourier transforms (based on sine/cosine) were derived, and just substitute in sinh/cosh.

If you were a formula plugger or just memorized facts, you'd be dead in the water.

I do think that's one of the reasons it's easier to do in physics. You're taught to see math as a language and therefore need to interpret it. With that in mind who cares if you memorize formulas and can churn out some algorithmic computation. You'll memorize formulas "accidentally" as you use them frequently. But if you don't know how to interpret the math you're completely fucked and frankly probably won't do well as a physicist. Much of the job is translating back and forth.

I actually loved my classical mechanics class. The professor was really good and in the homeworks he'd come up with creative problems. The hardest part was always starting. Once you could get the right setup then you could churn away like any other (maybe needing to know a few tricks here and there).

Coming over to CS I was a bit surprised how test based things were. I'm still surprised how everyone thinks you can test your program to prove its correctness. Or that people gravely misinterpret the previous sentence as "don't write tests" rather than "tests only say so much"

It's normal for young engineers to believe they can write code that cannot fail, design parts that cannot fail, design bridges that cannot fall down, etc. Fortunately, it was beaten into me in my first job that the idea is not to create designs that cannot fail, but to create designs that can tolerate failure. It's a very different mindset.

Unfortunately I don't think this is being beaten out of people these days. I meet plenty of people that are seniors at quite reputable companies that believe that and it scares me...

What I hear sometimes is along the lines of "I know my program has crashed, but I know it can continue to run safely."

What was your first job?

Designing machinery parts for the 757 airplane.

Reminds me of this article - https://firstthings.com/math-is-erotic/ - strangely titled “Math is Erotic” but talking about the relationship between Maxwells Equation and water waves, and magneticism.

Funny enough, I was going to include those relationships in my original comment but pulled them out for brevity. Was going to be something like "I think it is really hard to get a physics degree and not stumble upon the similarities of Gravity and Electromagnetism where there's a striking similarity in forms. I think most people start by thinking there's some link to the unification (and imagine solving a TOE) but you end up finding incredible beauty in something seemingly simple like the inverse square law." I actually remember one of my favorite lectures (and expanded on as a homework question) was deriving why everything is in inverse square law and not some variation like r^1.9 or r^1.5. It just completely changed how I saw these equations and even the process of analysis. I was a pure experimentalist before that class and gained a lifelong appreciation for theory because of it. I'm in ML now, so definitely still in application/experimentalist mindset but when I stumbled upon this Knuth quote it resonated strongly

  If you find that you're spending almost all your time on theory, start turning some attention to practical things; it will improve your theories. If you find that you're spending almost all your time on practice, start turning some attention to theoretical things; it will improve your practice.

Both are beyond my math, but I have found several more elementary cases where calculus shows simple relationships between related formulas.

And I recall a sci-fi short story long ago, technological civilization on a single continent with a permanently clouded sky. They had not figured out they were living on a sphere, they were having trouble with train tracks mysteriously being the wrong distance and train passengers feeling light on the high speed trains. I didn't check the guy's math but it sure seemed right when the answers looked exactly like Einstein's equations even though the units were very different. (Limiting velocity = orbital velocity, the discontinuity being weightlessness.)

That's basically a take-home at that point (assuming open book--or at least honesty) and, yes, you're now computing with classmates who will spend a weekend on it. It's the same problem as companies giving take-home interview problems that you should only spend an hour or two on.

At Caltech, exams were take-home, with a 2 hour time limit. It was on your honor to abide by the 2 hour rule. I used my alarm clock.

Ya know, the funny thing about students - if you presume they are honest, they tend to be honest. The students loved it, I loved it. If anyone cheated, the students would turn him in. Nobody ever bragged about cheating, 'cuz they would have been ostracized.

Besides, I actually wanted to learn the stuff.

> if you presume they are honest, they tend to be honest. The students loved it, I loved it. If anyone cheated, the students would turn him in. Nobody ever bragged about cheating, 'cuz they would have been ostracized.

I think if you look at the 2012 Harvard cheating scandal, it's clear that this isn't true. There, the professor presumed honest students, hundreds cheated, and no student reported.

https://en.wikipedia.org/wiki/2012_Harvard_cheating_scandal

It doesn't always work, that's for sure. I suspect one of the things the admissions committee did was try to filter out the cheaters. Explaining how the honor system worked was part of the freshman orientation camp (held on Catalina Island).

One reason it did work is the students liked being trusted, and they did not like anyone that would threaten the system, and would turn them in.

BTW, that was 50 years ago. I have no information on how the honor system is fairing today.

A couple of my upperclassmen professors used open-everything exams, notes, textbook, even the Internet was allowed. Although time was tight so if you felt like you had to Google something, you better not have to do it a second time.

My time was pre-intertoobs. My freshman year was the last year one saw slide rules and punch cards.

The ASR-33 teletype lasted another year.

I ceased knowing everything about my computer in the late 80s.

Why should students care how long it took for other students to perform the test? Is this because tests are graded on a curve? If so, that's the real problem

Reply on news.ycombinator.com