Race to the bottom: Red Grader, Blue Grader

[Khendraja'aro]

Yeah grading just is not that complicated. Why do you think students typically get similar grades from class to class? How is it that bright folks get 4.0 GPA in college? Because the control over the grade is mostly in the hands of the students. Yes some subjective bias exists but ultimately if you are proficient in whatever subject you are taking you will likely get a better grade.

Well, again, you don't know what you're talking about. For example (and this is just one example), teachers who were told that a particular pupil was a great performer almost automatically gave him good grades. While teachers who were told that the same pupil was a bad performer gave him bad grades... That's called the Halo effect.

Ah, and I see: Pupils getting constant grades automatically verifies the validity of grades. I'd love for you to present that particular argument to a professor in Statistics.

Lets flip this to the real world for a second. You have to rate 10 employees, you have challenges just like in grading. Do you throw up your hands and just mark pass fail and give high performers the same bonus/raise as an average performer? No that's retarded.

Well, what exactly do you mean by the word "performance"? How do you measure that? I mean, in Sales it would be pretty clear-cut: The profit you yielded for the company. However, you'd also have to define some kind of grading system there as well - I mean, what about the sales guy who got a foot in the door in one important market, but in order to do so, he had to offer a serious initial rebate?
And how do you measure performance in Techsupport? In Development? In Marketing?

Take Development: Do you use a moronic measurement like LoC (Line of Codes)? In that case, I could easily triple my LoC by simply replacing:

x = (question == "test") ? true : false
by

if question == "test
x = true
else
x = false
end

Or you could base it on the number of tests you're writing (yielding a lot of "assert true" tests) And so on. Which means that every kind of metric is flawed in some way. And if you don't recognize the flaws, you'll be doomed to deal with sub-par metrics which don't actually reflect reality.

[Nessus]

What I'm saying is your going way over the top over a single test.

Though I really don't get your point about "punishing" "sin" "failure" nonsense. Grades are designed to determine someone's proficiency. If you are not proficient, for whatever reason, it should be documented appropriately. You don't auto pass a dyslexic just because they have a harder time spelling. Spelling of course may be a bad example due to spell check but overall teachers should never help someone pass unless they actually have learned/demonstrated the subject.

Are you suggesting that we should give out better grades for some people because they are not as able as others?

The lack of comprehension displayed here doesn't really invite further discourse, as ideas aren't being exchanged. I would posit it is not 'nonsense' but perhaps in your case introspection has lead to a different theory of the conservative mind.

[Aimless]

I think we're losing track of the most important question:


why is it important for teachers to grade students on a five point scale?

Answering that question is key to understanding the madness on both sides of the issue, and being able to come to an answer that we can all agree on helps us define the problems associated with grading as well as to uncover possible solutions.

Any takers? Any suggestions besides "because it is"?

[Aimless]

If you are not proficient, for whatever reason, it should be documented appropriately.

Let's take a stab at identifying one of the problems. There it is, right there in your post. For whatever reason. At the same time, you're working with the a priori belief that the reason is that the kid is lazy, stupid, etc. In other discussions, you want to blame the teachers or blame the schools or even blame the parents. You can't have your cake and eat it too, Lewk.

[Khendraja'aro]

Well, it is because the people responsible believe that it yields some kind of meaningful metric.

I mean, it's even worse in Germany: We have numbers as grades.

Which means that they'll do averages. And a pupil getting a 2.3 average sounds oh-so scientific that it has to be true!

Nevermind that this would be a 2.3 ± 1 in any serious scientific publication (numbers made up on the spot, people!)

[Nessus]

Well, it is because the people responsible believe that it yields some kind of meaningful metric.

I mean, it's even worse in Germany: We have numbers as grades.

Which means that they'll do averages. And a pupil getting a 2.3 average sounds oh-so scientific that it has to be true!

Nevermind that this would be a 2.3 ± 1 in any serious scientific publication (numbers made up on the spot, people!)

How would one determine the error bars in the case of one student? Deviation across several semesters?

[Khendraja'aro]

How would one determine the error bars in the case of one student? Deviation across several semesters?

Well, first of all you'd have to actually quantify the errors. Then you'd have to define what a grade actually means. And then you could think about defining the error bars

[Nessus]

Yeah, that was my reasoning too, but since you're closer to the edumacation business then I am, figured it was worth to ask.

[Lewkowski]

Well, again, you don't know what you're talking about. For example (and this is just one example), teachers who were told that a particular pupil was a great performer almost automatically gave him good grades. While teachers who were told that the same pupil was a bad performer gave him bad grades... That's called the Halo effect.

Ah, and I see: Pupils getting constant grades automatically verifies the validity of grades. I'd love for you to present that particular argument to a professor in Statistics.



Well, what exactly do you mean by the word "performance"? How do you measure that? I mean, in Sales it would be pretty clear-cut: The profit you yielded for the company. However, you'd also have to define some kind of grading system there as well - I mean, what about the sales guy who got a foot in the door in one important market, but in order to do so, he had to offer a serious initial rebate?
And how do you measure performance in Techsupport? In Development? In Marketing?

Take Development: Do you use a moronic measurement like LoC (Line of Codes)? In that case, I could easily triple my LoC by simply replacing:

x = (question == "test") ? true : false
by

if question == "test
x = true
else
x = false
end

Or you could base it on the number of tests you're writing (yielding a lot of "assert true" tests) And so on. Which means that every kind of metric is flawed in some way. And if you don't recognize the flaws, you'll be doomed to deal with sub-par metrics which don't actually reflect reality.

Sure creating metrics that reflect reality is important. Just talking on a general level though, companies "rack and stack" employees. Employees are given "grades" so to speak and are given varying levels of bonus/raises and perks. This is true to one degree or another from menial labor in a fast food store to finance folks on wall street.

A measurement does not need to be perfect to be used. The contempt you have against grades is just odd.

[Lewkowski]

Let's take a stab at identifying one of the problems. There it is, right there in your post. For whatever reason. At the same time, you're working with the a priori belief that the reason is that the kid is lazy, stupid, etc. In other discussions, you want to blame the teachers or blame the schools or even blame the parents. You can't have your cake and eat it too, Lewk.

Each person is responsible for their own achievement and failings. However there is an influence factor.

Just like a criminal is 100% responsible for their crimes. However they are more likely to be a criminal if their parents did not teach them right and wrong. Does that absolve the criminal from his crimes? Of course not but we shouldn't put our hands in the sand and say that people do not have influence over others.

When it comes to teaching, incompetent teachers can waste students time, teach them incorrect things, fail to motivate and a whole host of other bad behaviors. This incidentally is why its important to not fully rely on grades only because teachers can be subjective. Objective tests administered by a third party is the best way to determine if learning has occurred.

[Lewkowski]

I think we're losing track of the most important question:


why is it important for teachers to grade students on a five point scale?

Answering that question is key to understanding the madness on both sides of the issue, and being able to come to an answer that we can all agree on helps us define the problems associated with grading as well as to uncover possible solutions.

Any takers? Any suggestions besides "because it is"?

A five point scale is more accurate than a 2 point scale (pass/fail). A 100 point scale is even more accurate. Those 100 points those fit into categories such as (90 to 100 = A).

[Aimless]

The compulsion to use a bad ranking tool just because it's easy reflects a deep-seated belief that ranking people is so important, in and of itself, that it must be done by any means possible. Where does this belief--that ranking is important for it's own sake, so important that it's better to do it poorly and unfairly and with dire consequences than to not do it--come from?

[Khendraja'aro]

A five point scale is more accurate than a 2 point scale (pass/fail). A 100 point scale is even more accurate. Those 100 points those fit into categories such as (90 to 100 = A).

That's false. If your data is inaccurate you won't get more accuracy by increasing the scales.

Case in point: Physics. We had multimeters which were able to display currents with up to 8 digits behind the delimiter. Did that mean that we were actually able to capture data with such a precision? Nope.

Another example: You have the numbers 2, 4, 5, 2, 1, 3, 2, 1, 4, 3, 2 which yield an average of 2,6363... Does that mean that you have an average of infinite precision? Nope. Since your original data did not have any significant digits behind the delimiter, your average should reflect that fact.

Which means that your scale has to reflect the precision of your data. You could up your scale to a 1,000,000,000 point scale. Does that mean that you can actually measure that precise? Nope.

If your initial data points are rubbish, trying to do statistical nonsense like yours is like gilding a turd. It may look pretty - but it still is a turd at its core.

[Khendraja'aro]

Addendum: During the last three years at school, we were switched from the usual 1 to 6 (1 being best, 6 being worst) grades to a point scale ranging from 0 to 15 (0 being worst, 15 being best).

Even then, for one teacher that did not suffice and he had to differentiate further and divided every point into three additional niveaus (like, 13+, 13 and 13-).

We always puzzled about what a 0- was to mean. Maybe: "Would have been better if he hadn't attended classes"?

[Wraith]

I mean, it's even worse in Germany: We have numbers as grades.

Our grades actually have numeric values. A=4.0, B=3.0, and so forth. What to do with +/-'s, if they exist, varies, but the base grade per letter is constant.

More on topic, while the grade scale does have it's problems, especially in more subjective subjects, pass/fail has it's own set of problems, and doesn't even fully address the problems with the grade scale. Subjective grading can still impact the outcome, and you lose information on whether a student had only barely passed a subject, or whether they demonstrated complete mastery of it.

There are some groups that try to counter the subjectiveness of letter grades by finding a more relative ranking, which makes some sense since everyone's grades can be expected to skew in the same way for many of the problems/modifiers in a given education system. In a case like this, the professor who gives a wide spread of grades versus the professor who gives a narrower spread won't matter. Grading on a curve is a way to translate relative grading into the letter scale, and the class rankings is the way to do the reverse. However, this of course has it's own issues. If you're in a class full of geniuses, a low ranking/curved grade is still probably something to be proud of, but it won't really look like it.

IMO, a five point absolute grading scale is preferable, because you don't lose as much information as in pass/fail, it minimizes the misinformation that could creep in through a more discrete scale, and it judges the students on their own competency alone. YMMV, and I doubt I'd convince anyone with different preferences to switch.

tl/dr: All grading systems have their problems.

[Loki]

I had a professor who'd give A-- and B++. Though that's mostly a matter of grade inflation at grad school (B is pass, which means anything less than an A- is considered a bad grade).

By the way, as someone who's graded 6 semesters worth of classes, I could tell you that it really isn't very hard to give accurate grades at the bottom and upper tiers. It's very obvious what an A paper (or exam) is, just like it's obvious what's a C, a D, or an F. The only problems I ever have is differentiating between B-/B, B/B+, B+/A-. Even then, there's a clear difference between a B and an A-. And I've graded for classes that emphasized papers, exams using multiple choice, and exams using essays.

Edit: I think the real problem isn't comparing an A- to a B+ from the same grader; it's being able to compare grades given by different graders.

[LittleFuzzy]

. . . reflects a deep-seated belief that ranking people is so important, in and of itself, that it must be done by any means possible. Where does this belief . . . come from?

The ability to engage in higher-order thinking? What else do you think our brains are doing all day *and all night*? They're processing, classifying, and archiving.

[Lewkowski]

That's false. If your data is inaccurate you won't get more accuracy by increasing the scales.

Case in point: Physics. We had multimeters which were able to display currents with up to 8 digits behind the delimiter. Did that mean that we were actually able to capture data with such a precision? Nope.

Another example: You have the numbers 2, 4, 5, 2, 1, 3, 2, 1, 4, 3, 2 which yield an average of 2,6363... Does that mean that you have an average of infinite precision? Nope. Since your original data did not have any significant digits behind the delimiter, your average should reflect that fact.

Which means that your scale has to reflect the precision of your data. You could up your scale to a 1,000,000,000 point scale. Does that mean that you can actually measure that precise? Nope.

If your initial data points are rubbish, trying to do statistical nonsense like yours is like gilding a turd. It may look pretty - but it still is a turd at its core.

If you pigeon hole three test scores into A B C D categories there will have to be rounding taking place. If someone has 3 A 1 B 1 C what does that give you? To use an American public school metric that could have been a 90, 90, 91, 80, and 75. Or it could have been a 100, 100, 100, 89, 79. Which is going to get you a more accurate measurement of student achievement?

As far as bad data that all depends on the test. If the grader can't grade something accurately (again not perfectly) why would you trust them to do pass/fail correctly?

[Lewkowski]

The compulsion to use a bad ranking tool just because it's easy reflects a deep-seated belief that ranking people is so important, in and of itself, that it must be done by any means possible. Where does this belief--that ranking is important for it's own sake, so important that it's better to do it poorly and unfairly and with dire consequences than to not do it--come from?

Because education should be a reflection of real world. And people are ranked. When you decide to give a promotion to someone you will look at all of the people who want the job and measure them against each other. Ditto for hiring, ditto for raises, ditto for choice assignments.

Furthermore grading allows someone to measure how much they have learned. Forget comparing for a second but if a child made a 100 on a test that is a huge difference then a 70. However in a pass/fail paradigm those kind of pin point grades could not be assigned.

[Nessus]

I love Poe's law.

[Aimless]

Because it's easier to differentiate between black and white than between a hundred shades of gray?

Consider your percentage scores. Do they say anything at all about the questions answered? Do they say anything reliable about the essays evaluated?

[Khendraja'aro]

If you pigeon hole three test scores into A B C D categories there will have to be rounding taking place. If someone has 3 A 1 B 1 C what does that give you? To use an American public school metric that could have been a 90, 90, 91, 80, and 75. Or it could have been a 100, 100, 100, 89, 79. Which is going to get you a more accurate measurement of student achievement?

As far as bad data that all depends on the test. If the grader can't grade something accurately (again not perfectly) why would you trust them to do pass/fail correctly?

I'll give you another example which should show another problem with averages:

Pupil one has the score progression 60 70 80 90
Pupil two has the score progression 90 80 70 60

Both will get the same average. However, would the average be really fitting in this case? I mean, pupil one is clearly improving while pupil two is clearly losing proficiency.

And you obviously did not understand my comment towards Loki. I did not say that pass/fail is superior in every way. I did say that in Loki's example it would effectively amount to the same thing.

Oh, and by the way: You are actually not allowed to average grades. Statistics plain out forbids it since they're actually ordinal numbers.

[Aimless]

The ability to engage in higher-order thinking? What else do you think our brains are doing all day *and all night*? They're processing, classifying, and archiving.

Hmmm. I think that this pronounced need to rank people is rooted in lower order thinking, in the genital area. I think it's from our monkey days. Or it could be a product of his upbringing. Or a product of extreme and unsettling uncertainty. It could be a lot of things. Swedes are notorious for being reluctant to rank people in many ways and I'm pretty sure their brains are about the same as Lewk's.

[Lewkowski]

Oh, and by the way: You are actually not allowed to average grades. Statistics plain out forbids it since they're actually ordinal numbers.

Yeah I don't know how you guys do it in your country but grades are averaged ALL the time here.

[Lewkowski]

Hmmm. I think that this pronounced need to rank people is rooted in lower order thinking, in the genital area. I think it's from our monkey days. Or it could be a product of his upbringing. Or a product of extreme and unsettling uncertainty. It could be a lot of things. Swedes are notorious for being reluctant to rank people in many ways and I'm pretty sure their brains are about the same as Lewk's.

A reluctance to rank people probably stems from some sad little fear of hurting people's feelings. Oh think of their self-esteem!

People are evaluated at their jobs. They are evaluated when they apply for credit. They are evaluated when they get an insurance quote. They are evaluated when they try to rent an apartment. Coddling people in school makes them ill prepared to deal with reality.

[Khendraja'aro]

Yeah I don't know how you guys do it in your country but grades are averaged ALL the time here.

Well, that's another sign that grades are misused if they're averaged. Again: Statistics plain out forbids the averaging of such numbers.

In order to do a meaningful average, a number scale has to fulfill two criteria:
It has to have a zero point. It also has to have equidistant points on the scale.

Both are NOT fulfilled by grades.

A reluctance to rank people probably stems from some sad little fear of hurting people's feelings. Oh think of their self-esteem!

People are evaluated at their jobs. They are evaluated when they apply for credit. They are evaluated when they get an insurance quote. They are evaluated when they try to rent an apartment. Coddling people in school makes them ill prepared to deal with reality.

Oh, great, here we go with the moronic "coddling" argument again. Where did I say that I do not want to give feedback to students?

I do think, however, that a "Has problems with Calculus, does great in Algebra" *) is a tad more meaningful than "C", don't you think?

*) Yes, it's a shortened version!

[Aimless]

Yeah I don't know how you guys do it in your country but grades are averaged ALL the time here.

Yes and creationism is fact, obama is a saudi prince, etc. Can you, even for a moment, entertain the notion that some things that Americans do all the time are in fact wrong? Like buying lottery tickets and houses and stuff.

[Aimless]

this may be relevant:

www.sonoma.edu/users/w/warmotha/grading.html

[Lewkowski]

I do think, however, that a "Has problems with Calculus, does great in Algebra" *) is a tad more meaningful than "C", don't you think?

How big of a problem with Calculus? A grade will tell the individual how far off the mark he is. A 50 is a huge difference from a 15.

In order to do a meaningful average, a number scale has to fulfill two criteria:
It has to have a zero point. It also has to have equidistant points on the scale.

I'll say it again, your dumb. If there are three 10 question tests. I answer 7 correctly on one. That is a 70. Another one I answered 9 correctly. That is a 90. And yet another one I answered 8 correctly. That is a 80. I average those three scores together and I get an 80. I've made a B. This is done in pretty much every grade from 1 to 12 and most colleges as well. Your nonsense about blah blah blah not statistically meaningful is just that, blah blah blah.

[Nessus]

I'll say it again, your dumb. If there are three 10 question tests. I answer 7 correctly on one. That is a 70. Another one I answered 9 correctly. That is a 90. And yet another one I answered 8 correctly. That is a 80. I average those three scores together and I get an 80. I've made a B. This is done in pretty much every grade from 1 to 12 and most colleges as well. Your nonsense about blah blah blah not statistically meaningful is just that, blah blah blah.

No one is this stupid; this is a blatant troll.