Hacker News Reader

> old school grading was often somewhat capriciously difficult and arbitrary

Yes! The people complaining about grade inflation want to use grades to compare students across classrooms, yet the level of achievement indicated by a grade is very much specific to a particular class, peer-group, and professor.

Grades should never have been used for cross-classroom comparisons in the first place and if market dynamics have inflated them to the point where it is impossible to continue to abuse them in this manner, I'd call it a good riddance!

That is the problem with metrics. Once there is a metric, a grade or a citation count or an impact factor, people will use it for doing comparisons.

For example in my home country, a relatively big EU member, CS and other technical schools used to give pretty poor grades. Each year, the best CS student among all schools would graduate with a GPA only slightly over 3 (B).

If you applied to a PhD program say at MIT or Cambridge, this would put you in a pretty bad position unless the committee had a lot of insight on local grading policies or you had exceptional achievements to compensate. Not fun.

No metric is free from this problem, but grades have a much more severe case of it than competing metrics (citations, standardized tests, competitions), and we should treat them accordingly. Even if all metrics are bad, we should try to favor the ones that are less bad.

Elite admissions all claim to control for grade inflation/deflation, but since they don't divulge details they almost certainly do a poor job of it. I'm not at all sure it's possible to do good job of it, but some of the obvious easy "fixes" (e.g. using percentiles) would at least alleviate the problem you describe.

Isn't that an argument for grading on a curve though? Then MIT knows you're in the Xth percentile even if completely unfamiliar with the programme.

Not really? You'd have to normalize across _all_ applicants, otherwise you're putting applicants from countries with a harder curriculum at a disadvantage. 98th percentile from X doesn't mean the same as 98th percentile from Y.

But X and Y universities are being compared on prestige anyway, so the ancient and revered Y's 98th percentile has already beaten X's 100th.

I'd have thought (with no admissions experience) that 98th percentile at an unheard of university is more meaningful than 4.3 points out of 5 from the same.

> For example in my home country, a relatively big EU member, CS and other technical schools used to give pretty poor grades. Each year, the best CS student among all schools would graduate with a GPA only slightly over 3 (B).

That is why there are tools to convert grades from different countries into US letter grades: https://applications.wes.org/country-resources/resources.asp

My chem department prided itself in the fact that it's core courses and grading system hasn't been touched since around 1995, when they built the current online lab submission website. "You would have gotten the same grade two decades ago," they promised.

Once you were handed the exams, it was obvious why this was the case. They were hard as shit, class averages would hover around 55%, curved to a C even.

We would later be shown the averages of other sections, lecturers, and years, and they would never sway more than 2-3% off the median score despite different sections, professors, decades even. The only constant maintaining this consistency was the content and pace of the lectures and exams, and the grading scale. Despite all the intergenerational turmoil and educational upheaval debated over the past 25 years, here was the chemistry department throwing perfect darts.

Schools need to decide how rigorous they should be, establish a very clear grading system that tests that rigor, and stick with it for decades.

If you develop a rigorous, reliable assessment for a rigorous, reliable course then the very last thing you should want is to brand its scores in a way that invites unfavorable comparisons to degree mill rubber stamp "A"s by people who don't know or can't be bothered, yet this is exactly what you do by calling these grades, fitting them to the ABC scale, putting them on report cards, and weighing them into GPAs. It's madness.

Such an assessment would have much more value as a standardized test, entrance exam, etc. Yes, those are much harder to develop, but that's the nature of the beast. If you want to facilitate inter-institution comparisons you need to develop inter-institution standards.

There's an orthogonal issue here in assessing the assessment. Imagine a terrible test, one with no meaningful relationship between questions and answers. The scores will follow a very reliable, consistent bell curve with very reliable, consistent statistics. So we can't actually take that as an indication of a good test, even if the best possible test would also behave that way.

The number of people that will say, without irony, that they expect upper level classes to follow a normal distirbution in grades boggles my mind. College by definition involves a great deal of gatekeeping and self selection.

As for the larger trend, I can't help but think it's as much due to better teaching and exponentially more available resources (both for self study ie internet and provided by universities (got to spend that tuition money on something)). From my limited experience (parents went to the same school I did) on a material level classes are objectively harder (more information, deeper coverage, more advanced topics) so sometimes it feels like the "grade inflation panic" is built on a older generation upset that young people might be outpacing them.

Here we don't usually grade on a curve (just move the mapping of points to grades a little), and a general observation seems to be that there usually is an overlap of two normal distributions - the ones that understood the material and the ones that didn't.

> a general observation seems to be that there usually is an overlap of two normal distributions - the ones that understood the material and the ones that didn't.

For any sufficiently large class with challenging material, this has been my experience as well.

You either learn the material and get the questions on the tests correct, complete projects that work, write essays that make sense, etc.. or you do not.

In the case of writing & humanities often the problem is that it is a lot more difficult to tell a correct answer from an incorrect answer. It could be as subtle as the students writing not matching with the professor's political bent.

But for STEM it is way more clear. Wrong answers on a math test are wrong. Software projects that don't compile should fail.

Not sure what the problem is. If the whole class is failing either the school is letting in students who don't belong, whoever taught the prerequisite classes failed to teach the students and passed them on with a curve, or the professor just didn't teach well.

In my mind it's more nuanced than that. An answer can be mostly right but maybe someone missed a negative sign or misread their parentheses and thus failed to complete the question correctly but still understood how the question was to be completed. Should they be penalized as much as a totally blank answer? Probably not. They've demonstrated some learning and they're almost as competent as someone who didn't make the same error. And it's not an error that the course is meant to correct or teach.

That is for sure a thing but that's definitely separate from grade inflation.

The grade inflating teacher might give the person who almost got it right an A, the teacher not inflating might give that student a B.

Even in these super difficult engineering classes at ultra hard schools there are always some people getting the answers right and showing all the work correctly so everyone knows they aren't cheating.

I had some classes that were incredibly difficult where realistically maybe a bunch more of us should have failed.. the teacher scaled things up but we sure as hell did not get undeserved As.

> But for STEM it is way more clear. Wrong answers on a math test are wrong. Software projects that don't compile should fail.

How hard should the problem set be? How long should students have to complete it?

As you increase the difficulty you will push down the score distribution. How do you propose to square the existence of a score distribution against the quite subjective level of "difficulty" of the work that you're setting for grading?

> Wrong answers on a math test are wrong. Software projects that don't compile should fail.

As a former grader and TA for both math and physics, I can tell you this is not the case. I've had students get the right answer the wrong way, stumbling into the correct numeric answer with invalid or incorrect math. I've also had students get 90% of the way through a large problem only to make a simple arithmetic error that lead to the wrong numeric answer.

Which student deserves more partial credit? Or should I pass the student who made math errors but got the right answer, and fail the student who grasped the concepts but made a trivial calculation mistake?

According to my math and physics profs, the student who got the right answer the wrong way would get 100% credit, the student who didn't would get 0% if it's multiple choice (often with those common math errors from years past intentionally given as potential options), and maybe 50% or so if it was a long form answer and the class was merciful. For those long form answers, we would be given explanations on exactly how they would be graded by the lecturer and later the TA in recitation before the exam. For ochem synthesis problems, the prof would give you credit if you reached the correct intermediates at least, even if you couldn't solve the entire reaction, or the TA would take a point off for every mistake, and if the question was weighted for 8 points and you forgot 9 things, you'd get a fair zero.

These aren't shades of gray, they are just protocols you need to establish and stick with before these questions come up during grading. Then you need to stick with these protocols as department policy for decades, so your department is able to maintain the grading scale no matter who the faculty will be 50 years from now. That's the sign of a good department.

I've never understood the use of "curves". Secondary ed doesn't use them as far as my experience goes. Intuitively, it seems like they're harmful because they leave students off-kilter, not knowing whether or not they understand the material. Confidence, or at least not being nervous, is such a key in learning.

What purpose are they perceived to serve?

Curves make your grade less dependent upon the difficulty of the tests, but more dependent upon the academic ability of your cohort.

The first thing it does is make it easier on a teacher. Making a test where most people won't get 100% and then curving is much easier than making a test where most people fall into the 60-100% range.

I had a midterm on which the second highest grade was in the 40s. The very junior professor certainly would not have been allowed to fail all but 1 student.

Curves sound like a tool for damage control while professors learn how to teach. It seems like it would be better to require professors to have a basic certification in Education so they'd know how to teach better. Teaching well is something you just can't pick up easily on the fly.

Professors not knowing how to teach is common for research universities in the US. Their main job is research, and teaching is just something they do on the side.

Yes, my understanding is the same except that professors are usually required to teach. My best professor in grad school, in terms of teaching ability, had a wife who was a high school teacher. He was head and shoulders above everyone else in the department. He clearly learned methods of teaching from her. He also had active research projects.

I'm a software engineer. I also coach girls basketball at the youth level in my after-work time. I read about teaching methods, how people learn, and I talk with teachers whenever I can. Often they can help me get difficult lessons across to the girls.

It's fine for professors to be great researchers. But, if they have to teach as part of it, they should strive to be good at it.

On every 55% average exam I've had, there would still be 1 or two students with near perfect scores. A bell curve is better to grade vs. one where you let everyone get an 85-90% and chop off the tail to the right of your curve. An ochem prof's job isn't to make you into a walking ochem textbook, it's to challenge you to think about problem solving using what you've learned from studying ochem, it should be hard. It really doesn't matter if you got a 55% anyway if the curve puts you at a B.

> My experience was that the trickier part was dealing with new freshman, who didn't always seem as prepared as they should be for college, especially with writing.

I went to a top 40 undergrad. Not stellar but I would have assumed reasonably high.

I was astounded at the poor quality of people’s writing ability. Most people just don’t understand the concept of stating an assertion and defending it with premises.

> Most people just don’t understand the concept of stating an assertion and defending it with premises.

Did you perhaps mean: defending the assertion with arguments built upon explicitly stated premises?

I’m all for pedantry here but I’m not sure this one makes a difference. From my understanding a premise is a statement from which you draw a larger conclusion.

You’re saying you draw arguments from premises to support assertions... but as far as I can tell the formation of the sub argument is just making an assertion supported by premises to begin with, the formation of which acts as a premise to support the larger assertion. So I’m not sure I see the difference

If I'd limited my writing practice and technique to what was required and taught in K-12 school I'd have been unable to do that, too. It seems ridiculous but arguing on the Internet taught me how to write lean, clear arguments that could stand up to being kicked around a bit. Not school. School just taught me how to pad papers to hit a word count and conform to silly "essay" formats. What was taught in middle and high school writing wasn't even useful for college, let alone crafting text that anyone who's not being paid to read your work would ever choose to pick up.

It's like attending a shop class that uses made-up tools and materials that don't work very well, don't exist outside shop classes, and aren't even very similar to their counterparts that do. Then the assignments all involve making things that seemingly deliberately avoid teaching skills useful to building items anyone would ever want, for any purpose—the analogy's getting away from me here, but you get the point.

I do think part of the problem is that good writing sense is, in part, developed by reading, and most kids never really pick up the reading bug and so don't have much hope of ever writing half-decently to begin with. But the classes, at least when I was in school, weren't doing anything for those of us who'd met that pre-requisite, either. I'm not sure who or what they're for.

I think this is an insightful comment, and probably matches my experience. But I’m talking about writing skills for the same kinds of typical class assignments, so I’m not sure it applies.