2

I know some advantages and disadvantages of using absolute grading as well as curve grading. Next semester, I plan to combine both by first grading absolutely, and then use curve grading to add points to the students.

For example, if one gets 80% (usually an A) absolutely, but when curved he gets a B, then I will use only the absolute letter which is the A. However, if one gets 70% (B) absolutely but when curved he gets an A, then I will use the A.

My reasoning is just like those who support curve grading. I might prepare too difficult exam tests or some things might not go according to plan during my course. While I actually support absolute grading, I think this might be a good idea. (?) I don't like the idea of just purely using curve grading that can lower high grades, or purely using absolute grading which is not always appropriate as university students' and instructors' performance are not absolute.

Of course, evaluating each course will always be the top priority, but this grading I hope can provide some small "healing" to error-prone college courses each semester.

ADDED:

  1. The curve I use is the usual bell curve for distributing A, B, C, D, and F..

  2. I am teaching mathematics for undergraduate.

  3. There are less than 30 students each batch.

  4. This is a new program in this university..and there are only 2 batches..so "past grades" are not recorded just yet unless I take data from other universities..which might not be a good idea?

All the inputs are very helpful. I do not plan to strictly use curve grading as I mentioned. I also believe the longer this program goes the better we can get..

  • 3
    Say a bit about the scale. Do you have 30 students or 300 in a class? – Buffy Dec 22 '18 at 12:14
  • Ill add to my post. – bms Dec 23 '18 at 10:32
7

You have assumed that students' performance follows a normal distribution, with a grade of C in the middle of that bell curve and very few grades of A or F at the tails. My own experience, and that of my peers, is that a typical college class has a cohort of good students who will mainly earn grades of A or B, some number of poor students who will earn unsatisfactory grades, and a few "in the middle" students. That gives a bimodal distribution, with the valley just about where the peak is in that bell curve.

I urge you to look at past classes and check the grade distributions when you have used absolute grading. If you don't already have something very like a normal distribution, forcing one will distort the assessment results badly.

Your question says your goal is to help students, and the problem is that assessments may be too difficult or the course didn't go as planned.

I'd like to suggest a different approach that's a little more work than artificially distorting the grade distribution, but is much easier to explain to students. (Um, and to department chairs, if it comes to that.) You haven't mentioned your field, nor whether your students are undergraduates, but if your examinations can have a reasonable number of questions, measure the performance on every question and then exclude from the computed grade any where performance of the class as a whole is below some cutoff you establish heuristically. Almost everyone's grade will go up, and you can easily explain that you excluded those questions or problems that were "too hard." (In the rare case where a student gets only the "hard" problems right, you can make an independent adjustment; that never happened to me.) I do this with a spreadsheet that has student names in the left column, then one column for each question.

Once you've evaluated an assessment, you must look at the questions you excluded as "too hard" and If the difficulty was with the question, fix it. If you believe the question is reasonable, then you need to review the material with the class again. When that happened to me, I'd warn the students that they would see that question or problem again, and so they should prepare for it.

If you do this for every assessment, you will find that your exam questions and your coverage of the material converge, and after a very few semesters, you will no longer have questions that are "too hard" and you'll be comfortable with absolute grading.

  • I'll note that automated grading systems or scan analysis will probably do a per-item analysis for you. Writing a program yourself to do it isn't that hard. Even spreadsheet macros can do the job. – Buffy Dec 22 '18 at 12:23
  • @Buffy Absolutely. I used a spreadsheet "template" with the formulas set up beforehand. – Bob Brown Dec 22 '18 at 13:12
  • I'll also note that there is a statistical measure (I forget the name) that correlates overall performance on an exam with performance on a particular question. If the correlation is negative, it is a bad question, since the "better" students tended to do worse on this question and vice versa. Maybe someone can supply the name and point to the formula for it. – Buffy Dec 22 '18 at 13:20
  • @Buffy: That's literally just correlation, or more officially, the Pearson correlation coefficient r. N.B.: The Blackboard LMS obscures this by using the nonstandard term "Discrimination" (explained halfway down this page). – Daniel R. Collins Dec 22 '18 at 20:09
  • 1
    @JeffreyJWeimer The goal is to lean not to write questions that are "too hard," while not penalizing students when it happens. See the last sentence of my answer. – Bob Brown Dec 23 '18 at 2:58
5

Actually, I believe that strict curve grading is both immoral/unethical and a misuse of statistics. It is immoral since the grade a student gets isn't dependent on only their own efforts but the efforts of others and in some ways just the current composition of a class, which is partly determined by sheer luck.

It is a misuse of statistics, since a distribution, such as a normal distribution, is intended as a derived or emergent property of the characteristics of large population. But when use in strict curve grading, it is imposing a distribution on a small population for which you have little or no evidence of its appropriateness. See the answer of Bob Brown for a further discussion of distributions that are likely, but not guaranteed to occur. Suppose your Physics class is composed of 30 Einstein clones, for example.

Strict Absolute Grading is also immoral/unethical (but perhaps less so) because it doesn't admit the fact that students are humans and humans make mistakes. Some of those mistakes are inconsequential, but absolute grading can make them very consequential. Also, in most uses, it doesn't reflect the fact that humans can learn from their mistakes. So, it turns the classroom into a grading scheme as opposed to a learning environment. The other problem with absolute grading, though not an ethical one, is the constant whining of students who have "just missed" the next mark. You would like to have a system in which that isn't necessary.

The other question that arises is what the grades are based on. I find that over reliance on exams, especially high risk exams is also counterproductive educationally. Among other things, they can force the students into harmful practices such as cramming. It is difficult, though not impossible, to create an exam for which cramming is completely ineffective, but such things can also be very difficult to evaluate. If an exam only evaluates objective facts, it has, in my view, little educational value and can be actively destructive.

I prefer, then, to base grading on projects and writings rather than exams and tend to favor group work, though that is an artifact of my field. I also try, as much as possible to think of myself as a teacher, not an evaluator. The educational process should dominate the whole system with evaluation being useful to learning, not just to classification into grade buckets.

One way to do this is to permit (and encourage) re-work by students after grading and feedback is given. The underlying scoring rubric is like that of absolute grading, but the first evaluation isn't necessarily the final one. I've even asked students to resubmit work after I've discussed proper solutions with the class. A re-submission can earn additional points/marks, but not up to full marks, so as to discourage too much dependence on re-work. There is no whining when students just fall a bit short. They can make corrections. Even correcting and resubmitting inconsequential mistakes is an advantage as it reinforces the learning in their minds.

This resubmission system need not be overly burdensome on the grader if students turn in the old work (along with the grader comments made there) along with the new, and marking (highlighting) the changes. A glance at the package can be enough to reassign grades for the new work.

Note that what happens when you do this is that you spend more time and effort with those students who most depend on your teaching, rather than those who nearly always turn in essentially perfect work.

  • Resubmission is a great solution indeed. The issue which you also mentioned, tho, is concerning the time. I don't mind the effort honestly, but here we are challenged with conducting research too..which is time-consuming. Therefore, we usually give only one additional test for some nearly failed students after the final exam..which is not common since it depends on each lecturer who has different research load.. I usually give it, but now I plan to give the test after this idea is done, perhaps? – bms Dec 23 '18 at 11:01
3

Please do not plan on using a curve.

Combining absolute and curved grading makes no sense. Curves should be used when a test or assignment was surprisingly hard. You should avoid curves when possible.

I might prepare too difficult exam tests or some things might not go according to plan during my course.

Based on previous exams, you should be able to calibrate future ones pretty well. I am now convinced that even if I posted the actual test with the answers beforehand, people would still fail it. Good students will study and make good grades, bad students won't, no matter how much hand-holding you give them.

Here are some alternatives to grading on a curve that will give students a chance to make a better grade.

Allow students to redo problems they missed on the exam for 25%-50% of the points

So a 10 point question is now worth 2.5 or 5. This makes students do some work on their own, but given others got the exam right, it should be pretty straightforward. This lets people possibly jump one letter grade (usually). For added security limit the change to say 15 or 20 points total.

Assign "bonus-points" homework assignments

Let students earn a set amount of extra credit that will be applied to their grade at the end. This has the advantage that students that may be working part/full time, or with a high course load can plan in extra time up front.

Allow students to drop one test

Drop every student's lowest test grade. Some professors don't allow people to drop the final, but I'd encourage you to allow students to drop the final if they have an A. Very few students change their end-of-course grade by the final, and most A and B student will generally do well on it.

This has the advantage that if one test is extremely difficult, it will just be the one most people drop. This is essentially a built-in curve but has two advantages

1) A curve pit students against each other - this doesn't. This allows every student 1 "do-over" if they had a bad schedule or an emergency. Great students are essentially rewarded by not having to study for the last test.

2) From a professors perspective, if one test is just super difficult, you've already built in a mechanism to fix it. The "super hard test" will get dropped.

  • Also, I'll point out that dropping n tests is conveniently built into existing LMS's. – Daniel R. Collins Dec 23 '18 at 4:02
  • Yes..especially that dropping test part, I considered it too. I believe what you mean by hard is that the question is not reasonable based on the learning outcomes, isn't it? The problem here is also that my department never allows too many students to fail.. When students do not reach the standards, they add points..which is something I won't do.. Hence, as you pointed in the first paragraph, isn't curve (not strictly) a solution? – bms Dec 23 '18 at 11:03
  • (con't) as it might not be possible due to the pressure of the deparment.. – bms Dec 23 '18 at 11:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.