Hacker News Reader

Immediately, you've come up with a solution that wouldn't fit my type of learning, and my comfort level. I absolutely abhor "final" exams. It's like a grand-finale that is make or break, forcing you to study a ridiculous amount of content and make sure you know/understand all of it for a single event in the future. If you fail, you have to do it all over again. No course-corrections, no motivational reminders/nudges. Nothing.

The example you cite in the beginning, your fear of the impending results of a small test. Now multiply that by 20 (as you said yours was just 5% of the final mark), and that is how anxious I would be for the entire course leading up to the final exam.

So drop the ubertests. Test students understanding using application and insight, not giant knowledge dumps that can only be tested in one long-winded sitting. (I exaggerate a little, sorry about that).

I have to wonder, if the outcome we are interested in is whether students understand the material, can't we do both? Some optional mid point tests, where the student can chose to keep the grade or bet it all on that final assessment?

I mean tests only exist to convince us that they know the material. Why does it matter how it works, beyond the test being an accurate reflection of their understanding?

Why not both? Some of my favourite classes have said, "Highest of midterms + assignment marks or final exam mark."

Now that is brilliant

Almost all of my computer science classes had two grading schemes - one "normal" one and another which was basically just taking your grade on the final exam and not counting anything else. That latter option actually saved me one time, but that's because I was astronomically lazy in that class and didn't do anything (didn't even show up to some quizzes, I think, when two quizzes added up to an exam). I definitely think that giving students the option is good, though, because if you asked me which I'd want I will always choose the small incremental tests over one huge test.

I always thought the time requirement was the breaking issue. If you could decide when you are ready for the test, would it matter which test style was used?

It's forcing students to learn on a fixed schedule that causes both issues. If they could go faster on easier topics and slower on harder ones, all this pressure goes away. You still have to learn it, but you take then test when you are ready, whenever that is.

Whether this is with small quizzes or with a final exam, I think the ability to practice and be sure you are ready would take away all the worries.

Of course, the incentive to go faster would be that you graduate sooner. But you want good grades, so you go slower. With guidance any student should be able to find a good pace.

> I do not think the pressure comes directly from the need for higher grades.

I strongly disagree, it entirely comes from the narrative instilled in all of us that we're screwed if we don't get top grades, that I am a huge failure if I can't get into an elite program. If that pressure wasn't there, neither would this problem. Some kids literally think it's the end of their life if they don't get into x,y,z school. There's no one telling them that college doesn't actually matter a fraction as much as it's built up in their head. Just get a degree, but I mean, at this point, 5 years out of college, you're not limited in the slightest by where your degree came from, not in this industry, where what skills you have and your hustle dictates all (granted a degree from an elite program gives you a marginal boost, but still marginal). My fellow Ivy League grads would get blown out of the water by some of the people I work with.

That is completely untrue. It may be true for tech jobs which rely on hard skills, but for any sort of management jobs, graduate school admissions, banking jobs - virtually anything else - pedigree matters a lot.

The majority of tech jobs here in Europe will expect a degree in CS or something similar in the listings.

At least in the UK, it is definitely not like that.

Most IT-related jobs do not mention a degree as requirement, and of those that do, many have an "or equivalent experience" clause.

This is how I remember it to be when I was looking for a job 2 years ago, but I also made a quick search for "devops", "senior developer" and similar positions in London through indeed.co.uk just now to confirm it.

yes, but that's it. be it the best UNI in europe or worst in most cases doesn't have any impact at all, in worst is a very minor disadvantage that can be easily overcomed by showing skillset.

I'm not sure what parallel reality you're living in but what you're saying is not universally true. Here in India , it makes a huge difference where you went to college. Whether it comes to jobs , opportunities or even raising funding for your startup having that top college pedigree makes all the difference. Almost all the top startups are by alumni from the well known places.

Some companies won't even give you an interview if you're not from a top place. I myself sometimes regret not being able to get into a good place.

So , even I wish that what you said was true. But sadly it isn't.

Grades matter a lot , atleast as a signalling device if not anything else and kids are making the most logical choice among the options available to them and noone can blame them for it.

My experience has been the same as OP. It's not a parallel reality, its the US. It definitely does not matter where you go. If you work hard and and keep pushing yourself into more challenging positions, you'll be a highly competitive candidate for employment no matter what school you went to. Some may argue that those who get into good schools have a greater propensity to push themselves and achieve regardless of their situation - we hope that's true. But I have never in my experience seen such blindness toward graduates of lesser schools.

That hasn't been my experience. Students at top CS schools have much more access to recruiters at well known companies, who assign recruiters to particular schools. It's a lot harder to get an internship at such a company from a school without an assigned recruiter. Good CS schools also help students to get internships earlier in college, after freshman year.

I asked a professor in charge of admissions for a graduate program about their desired gpa. His response- 'What school did you go to?'.

Students at good schools don't have some innate willpower to push themselves harder- they have better opportunities available to them. That isn't to say that students at other colleges don't do just as well, it's just takes more effort.

I guess I'm biased from my personal experience. I started working on internships in high school. By the time I was in college, the better internships found themselves. I went to a good, but not top!, LAC but I still did well. Had my choice of grad schools based on my work - not my school. But I decided to drop out for other reasons. I have since had no issues making the 6 figs... So to anyone who tells me you need a top-college degree to get somewhere or make good money, I'd say there are probably other things at play. Now maybe I won't get a CTO position without a degree, but that is not being addressed here.

Seriously, if you feel like you are in an unfair position or want to talk about this stuff, send me an email. I love talking about this stuff. It's not about the school you go to - it's about what you do (have done) with your time.

If it mattered that much, I would not have been able to find jobs in US (and I did, in big name ones). My university is a shitty, unknown (but state run and legitimate) school in ex-USSR.

I have heard this about India and other countries as well. I think the original article is mainly talking about the United States though, and as others have said, the University you attend here matters very little for most jobs. The main exception is your first job after graduation.

Gotta disagree wholeheartedly with you there. If I have one big exam at the end, I will procrastinate and likely end up doing poorly. Instead, in my medical school, we are quizzed at least once a week - this week alone I have 2 quizzes and 2 tests. I failed yesterday's quiz, as I knew I would since was at a wedding on the weekend. But I will know the material for the test on Friday, and now I know what kind of questions to expect.

Weekly testing lets me figure out how I'm progressing, whether or not my study habits are working, and also provides a significant amount of opportunities to do well - that quiz I failed yesterday really does not in anyway hinder me from acing the course.

I took a class from Matt Might and I like how we were graded:

"Your grade is your tests grade or your projects grade, whichever is higher." http://matt.might.net/teaching/compilers/spring-2015/#gradin...

This allowed both the really good test takers to succeed and also those who did better on projects, without all the pressure.

...at the expense of people who are moderately good at both.

I believe the sentiment was more: I understand that you may have a bad day/week/month and do poorly on a project, but don't sweat it. If you really do know your stuff you can still prove it on the exams.

I think it is fair and better than purely weighted grades because it allows for momentary failure while still requiring knowledge of the material.

Not really. If you're moderately good at both then you'll get a similar score on both, and they will average out to about the same as both. So you'll still get the score you would've gotten before.

As opposed to everyone else who will get a higher score than they would've gotten before.

That means your grade will be worse than it would've been otherwise when compared to your peers (which is the only useful meaning of a grade).

People who are really good at one thing but not another are very useful in society, and this grading system would seem to reflect that fact.

That would seem to justify the commonly done separation of project courses from non-project courses. It's not a great idea to lump the two together, nearly always one or the other suffers if the project is non-trivial. But even then one can suffer, it's just spread out across all classes instead of the individual class. Project courses are tricky to integrate into a curriculum well.

i wouldn't say specialization in taking exams is a useful skill for society.

You can't do well at exams if you don't know the material.

It means your grade won't be artificially inflated by people who know the material but take tests badly.

If you're moderately good at both, in the combined grade you would have done... (wait for it) moderately.

So what's the problem?

I am absolutely against that exam culture so I like that suggestion a lot. The exam perspective is just completely artificial which has nothing to do with the real world. That project based approach mirrors the real world much better where you can compensate for almost anything just by rising the effort one is putting into something.

Completely agree with that. As someone who came from the British education system (one big exam end of year) to the American college system (a test every week), I quickly realized that in the US I had no time to read around the syllabus. If you're being tested every week, you Have to stick to plan. Essentially, it is "teaching to the test".

I much prefer the British way. You have a lot of time to explore your interests without the risk of jeopardizing your future, you are given optional exams to test your knowledge, and you can bring it all together at the end of the year with a final exam.

Yeah I actually advocate for the moderate form of what you're expressing.

My issue with single, high-worth exams was that sometimes I would screw up, have some misconception in my head, and if it wasn't cleared up until the exam I was done.

I think you should be examined on a semi-regularly basis where each exam is worth an equal amount of grade.

That way the effects of various externalities, like I was sick, had something bad happen that week, or just didn't learn the material as well as I thought could be smoothed out of a period of time.

There are things that can be done within exams to counter those issues. For instance, the exams I set are heavily scaled. There may be 200 points available to be earned across 15 short essay questions. The raw scores generally fall between 45-90, which I then scale to set an acceptable average. So you can bomb or even skip questions and still be top of the class.

Not much can be done if a student is sick on the day, but I would argue that multiple tests on multiple days only increases the likelihood of sickness during at least one test.

I have never understood scaling, exams are meant to provide employers and yourself with a rating of how well you know the subject, ideally in comparison to everyone else in the field. So if you fail, but everyone else fails too, you get an A. If yo do 'well' but a group of 5 out of 30 excel, you get a lower than representative grade (representative of your knowledge). How about making the questions worth points and the 100 points = 100%? 93+ = A, 80-93 = B etc. That's how all 3 of my undergrad degrees worked. I knew what I needed to do for a grade and damn everyone else.

Normalisation makes more sense over larger populations. If most people across a large group do much worse on a particular exam, is it more likely that everyone screwed up, or that the exam was unusually hard?

Universities I have attended have procedures to deal with those instances in which everyone performs equally badly. It is almost always an indication that something else is amiss. Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct.

Sure, in exceptional cases investigation is probably necessary. But there are still more marginal cases where one exam gets worse results than another.

>Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct.

The point is that it's just as unconvincing to say the student that got 55% is comparable to another student that got 55% in an exam where that was the lowest score. Exams are not perfectly comparable measures of absolute "mastery", so we have to use statistics to try and derive as much information as we can from their results.

If students in one cohort consistently perform worse, it's probably their fault, but if students in one exam perform worse, it's more likely the examiner's fault.

A lot of professors in my field (physics) see tests as a learning tool where one is presented with new things to be figured out. In my undergrad, exams were typically open-book and consisted of a few long, multipart problems. It does not make sense to scale such exams to 100%.

Anyway, by setting the expected mean to around 50%, you get a more accurate evaluation of people's understanding (by minimizing truncation at the limits).

How about a ton of exam opportunities that are entirely optional and are just for preparation, giving you time to understand the material with checks from the teacher, followed by two tests at the end, with the greater of the two scores being the one that decides your grade?

(1) That's a lot of work for the prof. (2) Grade inflation is part of the problem. I don't see how taking only the higher grades helps this.

Increased work for the prof should of course be balanced by decreasing work somewhere else. How many courses could have (some or all) video lectures recorded beforehand and assigned to watch as homework, thus saving the teacher lecture time and allowing the students who need extra help (and who otherwise would slow the whole class down by interrupting non-video lectures) to visit during office hours, or ask during a smaller lecture Q/A time slot of normal class, and point out exactly what parts of the recorded lecture they struggled to follow? And those who don't speak up their questions will get frequent feedback on their misconceptions anyway from the optional tests and determine if they should try harder or just drop the course.

As a prof, pre-recording video lectures doesn't save time -- it's a lot of work. And then you can't see the faces of students and react to their current state of understanding. Students "slowing down the class" are about 80% of the time providing the primary value of in-person instruction: they're creating the interaction that makes a lecture valuable.

What you suggest has essentially been done, with "book" substituted for "video". That's why all those profs wrote books -- they figured the students could read beforehand, thus saving the teacher lecture time and allowing students who need extra help to visit during office hours. That's the dream; video is simply a change in format. I point out to students often what parts of the book they should return to when they struggled to follow.

The most interesting question for me is what new technology adds or allows beyond old technologies. Analogize: what were the elements that changed from reading a serial in the paper to watching a show on TV to listening to a serial podcast to watching a YouTube series? As someone currently employed as a "prof who makes videos" I refuse to do any video unless I think it adds something beyond simply changing information format!

This is pretty much how some of the better known standardized testing regimes work - SATs and the British O and A level exams.

Generally, outside of some relatively garbage introductory classes, I found that my classes either required 2-3 papers or projects, a midterm exam and a final, or some combination of the above. The more bullshit courses, generally with the shittier professors, would impose a "class participation" component of between 5-10%, and do other garbage like clicker-based quizzes during lectures to try to get people to actually come to class, instead of just doing the reading, when exams were not scheduled.

A much better fix would be: don't count the two worst tests towards the average. This would allow you to screw up every now and then.

The logical extreme version is to take the median score.

You don't want a scheme where you know your course mark before the final, because then you have no incentive to try, and median allows this. If your marks are 75, 80, 80, 90, then regardless of what you get on the final, your mark will be 80.

Presuming that "trying" on the final is necessarily a good thing of itself.

Microtests should be bonuses; they should contribute to your final grade but you should be able to get A on the final exam even with 0 bonus points. So for a person that works during the year this alleviates the final exam pressure while allowing others that for any reason weren't scoring a lot to prove they mastered what was required.

This is how it worked at my University some 15 years ago (for some classes at least). The teacher would give us optional assignments that if completed would score against the final exam. ASFAIR completing all assignments successfully meant you didn't need to take the exam unless you aimed for a higher grade than pass.

Disagree. While I did well in the same system in law school, most people I know hated the one-and-done testing system. That said, you do have a point in situations where testing is practically constant (and affects the final grade) - there's probably a happy medium to be found somewhere...

I certainly remember many students with that view. After first year they started carefully arranging courses and boasted about how they would never have to write another exam ... until the bar. I think I had a final for every course but LRW and a bioethics class.

Yep, same here - paper classes were a real pain and always made the end of a semester rather unpleasant with deadlines looming. Just saying that system is a bit extreme, and I don't think it's universally applicable.

Oxford has no exams in the second year for many subjects (and no tests other than maybe college 'collections' at the start of each term which don't count for anything, but my college didn't even do those for my subject).

Personally I really liked it as it gave me a lot of time to spend trying to really understand my subject without exams and revision hanging over me. It also helped I think that our tutors didn't generally set or mark the exams when we did have them. It did put a lot of pressure on for finals though knowing everything was hanging on how you performed over the course of eight exams in a fortnight.

Cambridge does this as well and I can't say I was a fan. In mathematics, three years of learning ends up being evaluated through five three-hour exams administered over the course of a week. All it takes is one to fall ill for that week and all of a sudden you don't get a grade. The best case scenario is you can convince the examination board your illness contributed to your performance and they might award you a DDH (Deemed to have Deserved Honours) which no one will really understand.

I went to Oxford, and this actually happened to me. I got ill during exams, and there's no mercy. I sat in the exam halls sweating, unable to breathe properly, trying to hand-compute an analysis of variance calculation. And a bunch of other trivial stuff like singular value decomposition and other linear algebra riddles.

I actually think having one big exam at the end isn't great. There's so much on the line you end up "learning" things too quickly, in a way that's focused on previous exam questions rather than actually learning stuff. And the stuff is actually interesting.

Perhaps a pass/fail system would be better, spread out a bit. The problem with the system is that incremental gains in the exam result are favoured over learning that's hard to quantify.

Yeah this is just the professors being lazy... The grief of having to write more tests and be accountable!

The UK system (at least when I was doing it) at GCSE (14-16) and A-level (16-18) consists of smaller module exams, but students are allowed to retake the exams each time they're administered. So if they do well, it takes the pressure off their future exams, but it's not the end of the world if they don't. (EDIT: They're also standardised, which is pretty much necessary for this to work.)

I think it's a good balance between the massive short-term stress of big exams, and the consistent stress of smaller exams.

It think the pressure comes from globalization. The meme for success has shifted from a local scope to the global one: You are only really successful if you work around the globe in an international company. There are however, due to hierarchical centralization, there are very few positions in the global market, so you end up with a lot of people competing for a very low number of high-end jobs. That creates pressure, and in return, out of empathy teachers grade more benevolently. At the same time the education systems became more centralized too, leading to competition as well: Good teachers are those with successful students, leading also to grade inflation yet again.

I’m unsure whether this problem is solvable (e.g. with basic income or laws that penalize globalization), or whether it will solve itself with time.

Learning, understanding & doing: real value for self Testing: external value for others

The latter is useless in both the real world and for the student. Tests/grades aren't the only way to show progress.

(disclaimer: dad homeschooling a kid)

I think the calibration of quizzes and tests is already an area of research that is receiving academic attention, it's just that there's a gulf between research and practice.

I also agree with emphasizing tests more, since measuring homework is so noisy (tutors and parents double-checking homework), but I definitely don't think having one big test is the right way. One of the reasons why people might face test stress is by dealing with the stress at the last minute.

I just got out of public highschool a few years ago and it was fairly easy 99% of the time. Even the final exams didn't take serious studying. Any work that was "hard" was generally more tedious than challenging. It was an easy coast to get 90s. Maybe it's not high schools in general, just specific districts.

"So drop the microtests."

I disagree. Tests are part of the learning process.

The problem is when expectation is 90%+. When that happens, there's no margin to differentiate between minor mistakes and actual misunderstandings -- the signal is lost in the noise.

A good test will amplify the signal and attenuate the noise.

Formative or summative tests?

There's no rule that says every test has to count to your final grade.

Well I suppose that is the Oxbridge model.

And you're right. It allowed me to pursue extracurriculars (rowing) throughout the year and then adjust to final exam studying at the end of the year.

It would be better to set grades on a distribution rather than an absolute mark so you know that an A is the top 5% of the cohort that took the exam.