Hacker News Reader

Comment by uberman | Hacker News Reader

I'm kind of stunned by any defense of:

"As a teacher I hand out problems people can work on but will not provide the solutions."

That is completely BS and my degree is in Math and I labored through the same during my undergraduate degree. It is unacceptable in my opinion in any field and I hope that other fields like engineering and medicine don't take this elitist approach.

Ever try or even see a "Leet-Code" style question that stated the problem without providing example results? Ever try to help someone with a problem on Stack Overflow who can state their problem but will not provide example data? How do you ever know you got the correct result? You don't. Perhaps worse, you are confident in a "solution" that includes an "off by 1" bug.

View on HN

Some people believe the purpose of education is "weeding-out" weaker students akin to a kind of tournament, or a harsh initiation ritual that members of a profession go through. These kinds of rituals enhance loyalty/camaraderie among those who survived (sunk cost) while also minimizing the number of new entrants/competitors (aka gatekeeping).

People holding those views would believe it's kind of "unsporting" to help students or make it easier in any way, that it would defeat what they see as the purpose (that being, the difficulty itself).

I think some degree of "weeding out" is required. Just because you got into Fancy University College it doesn't mean you are automatically great at all academic endeavors. If you sail through Math 101 and 102, and run into a brick wall in Math 202, maybe you aren't as great at math as you think you are. Or maybe you need to work harder.

We tend to think of higher education as some sort of inalienable right to all peoples across the world, but it just isn't. Everybody has a ceiling. Higher education, and especially elite universities, should be for those whose ceiling is higher than the average bear.

That said, until you get into the later years of higher education, "weeding out" doesn't have hard and fast boundaries. Atypical thinkers may have a blind spot that needs a different teaching method, or possibly "departmental policy" is just garbage ideas ossified into departmental culture. It's always worth revisiting those Chesterton Fences to see if they need to be altered or moved, or even torn down.

High standards are quite distinct from the indiscriminate "weeding out" effect that's most often the outcome of bad teaching. If you make every effort to teach Math 202 in the most effective way but also expect stellar results from everyone, that's not the same as what the Princeton student experienced.

> Maybe you aren't as great at math as you think you are. Or maybe you need to work harder.

Sorry, but that's just silly. Extremely silly.

You don't weed students out through the <<teaching process>>.

You weed them out through exam results...

I agree and it amazes me. We are implicitly asking students who want to learn something outside of their area of focus to do so at a community college or for "no credit". Rather than risk their GPA on a course where the philosophy might be "it was hard for me so I'm going to make it hard for them".

The first day of my math undergraduate, in a lecture theater of probably 300 students, the professor said confidently and with joy rather than remorse.

"look at the person to your left now your right. They will fail out of this program."

The benefit of the community college is that teaching is the norm. Risk to GPA is a side issue.

Some problems should be given without solutions, because it will help a student identify when their use of solutions is becoming a crutch to help them solve problems. But as you point out, some problems need to come with solutions so a student can check their work. I assume this is a reason why so many textbooks have answers for every other question instead of answers for no questions or all questions.

With technology we can improve this by having an answer that is available only after the problem has been attempted, but the online answer checkers seem to introduce a host of other problems that make it unclear if there is any net advantage.

> Ever try or even see a "Leet-Code" style question that stated the problem without providing example results? [...] How do you ever know you got the correct result? You don't. Perhaps worse, you are confident in a "solution" that includes an "off by 1" bug.

Yes. Well, no--even worse. When I took intro CS in college many years ago we were given a problem with incorrect example results due to an off-by-one error in the code written by a TA. After many hours of banging my head against the desk and eventually working several cases out by hand on paper, I pointed this out to the professor. Rather than correct the error he emailed the class that the example results were only intended to demonstrate the formatting of the input and output, and should not be relied on to be correct.

I had a similar problem. Where we were supposed to do some sort of optimization (loop unrolling or something) to a function. We had done this several times throughout the course. But this time loop was slightly different (unintentionally) and I managed to construct some answer that I wasn't happy with and turned in the exam last. We got the papers back and of course mine was wrong in spite of producing correct results, while everybody who blinding applied the wrought memorized mechanization we had learned got it right. It took forever for that professor to understand how bad the question was.

> Ever try or even see a "Leet-Code" style question that stated the problem without providing example results?

This style of question asks the examinee to devise a procedure for producing those results, not to come up with the individual results themselves. If a student is given a bunch of systems of linear equations to solve, an answer key isn't especially useful since checking a candidate solution is pretty straightforward.

When doing my MSc, my class submitted long-form essays on several topics. I asked when we would get the marked papers back, but that was not available - some excuse that the marking scheme was technical. I was bemused.

A compromise approach would be to provide numerical solutions so students know they're correct, but will still require them to put effort into working on the problem. This way instructors won't be afraid that students will simply look at the solutions before solving the problem.

I will add though that there are courses that use the same graded assigments year after year (my OS161-based OS course comes to mind, where multiple universities basically teach the same content). As a result full solutions are all over Github, and the honest students who work hard on completing the problems on their own are penalized over those who look over the solutions first before writing any code.

In my maths BSc, I've never had this happen. I've seen it in textbooks, yes, but there I kind of understand it—it would otherwise just bloat the textbook, and many people use textbooks in class with a teacher anyway.

I've usually either had practice questions with solutions directly provided or assignments to hand in (usually this was optional). The solution to these assignments would be posted after the due date and of course, you would get your submission checked by a TA.

When you filter on a bell curve because fewer spaces exist next year you see this approach because being helped gives you an advantage over one of the other hundreds the professor cannot help directly.

It's part of the elite status colleges give out with degrees. Only the top 10% make it through is great advertising.

It still amazes me that CS students are not allowed to google when working on their coding assignments.

Same here. I personally feel that ability of successfully google should be something we test for!

What about other subjects, physics, literature? Should they be allowed to Google too?

Literature students are _definitely_ allowed to do research on their assignments.

As a physics student, there were some exams where I was able to "google," but almost every exam I took allowed me to bring some notes.

Yep, open book exams should be the norm.

There is such a thing as an 'open book exam' where you literally have access to the textbook in the exam.

The purpose of such exams is not to test your memorization of the subject, but how well you understand it. The questions are more challenging, and are explicitly designed to see if you can take the information in front of you and synthesize or interpret it.

If you don't already understand the material, having access to it in the exam will not help as much - when the questions are properly designed - as you can't just copy down a bunch of disconnected facts. Instead you have to build an answer that relies on those facts and the framework that surrounds them.

Open book != google.

Most my CS exams in college were open notes / open book, but Google was banned.

> The purpose of such exams is not to test your memorization of the subject, but how well you understand it. The questions are more challenging, and are explicitly designed to see if you can take the information in front of you and synthesize or interpret it.

The obvious difference is the existence of Stack Overflow. Synthesizing a correct answer from my hand written notes is a proof of understanding. Being able to google a question and having the answer produced for you is absolutely not a proof of understanding.

Its the difference between letting a math student have a formula sheet, so the don't need to memorize all the different formulas they need to apply, and giving them access to wolfram alpha. The later makes the exam an exercise in copy / paste.

Oh I agree that google is the ultimate open book!

Perhaps certain subjects are better suited to having broader questions. I still think it would be _possible_ to create CS questions that would not be helped by having access to a search engine.

As an analogy, it's a little like the difference between a unit test and an assembly (integration, etc) test. The unit test question is just a simple one like "When did this King die?" or "What is the area of this triangle?".

An assembly test is more like "What was the legacy of this King's reign?" or "What is the area of this shape that is the intersection of various other shapes?". I mean, Wolfram Alfa is only so powerful, right? Of course websites add features all the time, so it's always possible that the day the exam comes out they add a new "What is the area of this shape?" feature.

A big part of exams for math majors is proving stuff, not applying formulae. Wolfram|Alpha won't be able to help you there.

(That said, there's stuff like proofwiki or math.stackexchange, so unless you really spend time on crafting a novel problem, a student that is good at "researching" might be able to find a solution to a similar problem or to a substep of the proof, etc.)

How often do any of those subjects require instant perfect recall of knowledge/topic/concept with high stakes?

That's my criterion on whether external references should be allowed in tests. A surgeon should know their specialty as they won't have time to consult a reference during a surgery. A physicist will always have the time to refer to something when performing their work, as will an editor or teacher.

That's entirely different from saying they don't need to learn it however. If I drop a list of 100 formulas and equations on a person who doesn't know how to use them, they're useless. That background knowledge on what's available and how to use the tools you have is what the class should be teaching you.

Or as my physics prof used to say, "I want the engineer building the bridge I drive over to have looked up a reference on the static weight load of the materials selected, not pulling figures out of their head".

We were always allowed to use 1 standard sheet of paper, front & back, for our tests. Most of the time you could answer most questions without needing them beyond referring to a specific equation to make sure you weren't leaving off a constant (damn factors of 2....).

I graduated in 2018 from a state school in the USA.

No helping other students on assignments.

All exams are: closed book(this means no notes or textbooks with you during the exam) and done with a pencil and a BLANK white sheet of paper. I still managed a 3.75 CS GPA, but we were literally writing syntactically correct C and C++ on blank paper. Why oh why?

Early on before I had learned C++ well, I would get marked off something to the tune of -30 for missing a curly brace.

No one will miss that IRL when the IDE shows you the visual prompt that it's not there, OR you waste maybe a minute of your life if you forgot it when using VIM... -30 man oh man.

It depends on the content of the class. I had a class in Android programming in college where google was fair game because the API / libs were so large that google was a necessity to access all the documentation you need to write android apps.

My low level systems class was closed-book, no google, etc. At some point you have to learn something. Pointer manipulation is not something you should have to look up every time you do it.

Teachers prepare you for life, and life often hands you problems "but will not provide the solutions", eh?

Reply on news.ycombinator.com