Dear @Hashem, this is a very good thread that enables us sharing ideas how to distribute marks on the students. I like the 100 points scale rather than the letters system because it shows the exact mark and distinguish between the students in the same category.

Dear @Hashem, this is a very good thread that enables us sharing ideas how to distribute marks on the students. I like the 100 points scale rather than the letters system because it shows the exact mark and distinguish between the students in the same category.

Yes, in my institute, we aim at bell curve; so is it in the Aarhus University, Denmark where I teach a course every year. But it is not strictly enforced in my institute, and also in the danish university. There is broad guideline for this only.  If one has too  many good students, s/he may give better grades to more number of students. So it becomes a kind of binomial distribution.

I aleays prefer the 100 points scale rather than the A, B, C, D way of marking; there is no probability of  bias or increase in the students` marks.

Bell curve can be used for Students Learning evaluation but has certain limitations

• Subject evaluated is easy
• Subject content requires more support to understand and / or new.
• Evaluated regimental
• Evaluated with concession

In the above cases the curve will be skewed towards high/ low.

I do not use the bell curve, since I think, in our case does not hold! If the majority of student come unprepared then if you use the curve everybody gets a good mark and at the same time it is ignorant! 

Dear Colleagues,

If we use 100 point scale, are we weighing students achievement precisely?

I agree with Costas that a bell curve is escalating the student grades although they are not well reached the outcomes. The question is what is the advantages and limitation of each methods, taking into account cultural diversity, level of the students, and instructors? Can we reach to an optimal assessment criterion? 

I remember having to use a smoothing, Bell curve generating app on all students' grades at Melbourne Uni.

There were certainly some pros - the faculty lecturers were very different, some were lenient and some very demanding, and they graded accordingly , and this bias had to be addressed, and some definite cons - students knew that as long as they were better than the bottom 10%, they'd make the grade, and that made them work less hard than they would otherwise have .

I'm sure there must be an 'operating point' - an optimal system leading to the most learning while still preserving justice and balance - but I've never been able to quite determine it - whatever system is used will be imperfect.

You are right H Chris. Thank you for the clarification.

Thanks to Prof. Hashem Adnan Kilani for it’s really an interesting question. I am strongly agreed with the answers of Prof. Abedallah M Rababah, Prof. Debi S. Saini, Prof. Mahfuz Judeh & Prof.H Chris Ransford .

A Bell curve generating app? ... How awful! Think about the underlying philosophy and its consequences! And that on a continent where so much really progressive educational thinking and practice exists (see for instance here on ResearchGate: 21st century skills ...). But perhaps they have already withdrawn this app from their toolkit?

Instead of "smoothing out" (which is a statistical trick, nothing else), one should EITHER allow assessors to have differential judgmental opinions but to be absolutely explicit about it OR offer a app which has some sort of tolerance or lenience parameter which can be set individually or forced to a certain value.

The important message is to make lecturers aware of their tolerance or lenience, not to mask it out.

For instance, in my assessment system (cf. Vossen 2014), I have explicitly built in such parameters, which can be set between 0% and 100%. That's fair, if you tell your students before, and don't change it anymore after the begin of the course. Can be made dependent upon the sort of course or test given, or upon the policy of an institution. But doing it via statistical means, isn't really serious, IMHO.

https://www.researchgate.net/post/Which_skills_must_21st_century_teachers_have_to_promote_high_quality_learning?

I no longer teach but I always used the Bell Curve. In all institutions I taught it was up to the individual professor to decide but I personally felt it to be a more fair approach. But I always made it clear to my students what it meant and how it affects their grades. I never had any problems with that.

Allow me to frame the issue this way. The grading system is intended to assess the students' mastery of the content of a course. That depends on a exam design that can accurately do that. So fundamentally the question ought to be " is the exam a valid assessment of the content?" If the exam is a statistically valid assessment then one would be justified to make a grading scheme that produces a normal distribution of grades. In 45 years in this business I have never encountered such a course.

     This kind of conversation nearly always ends up sounding like the grading system is viewed simply as a means of creating a publicly defensible spread of grades. It would be far more helpful if the conversation was about how to produce well designed, well taught courses with appropriate prerequisites. Assessing courses like that is a more straightforward affair and the grade outcomes are often not normally distributed. Wouldn't it be lovely if most students had the opportunity to participate in courses like that. The outcome would often be a grade distribution strongly shifted upward.

Thanks Gopinath. R. ·

Thanks to Paul Hubert Vossen, Angela Stanton, and James Des Lauriers

I can see how much your answers are debatable. I agree that the learning outcomes are more important than the curve. I actually, don’t mind if most of the students score hi reflecting that they master the content and the quality of their production.

Further to my previous considerations and recommendations. If the Bell Curve Smoothing is used with a grading system, this aggravates the problem enormously.

Therefore the first thing to do when getting into the business of manipulating raw judgmental data is abstracting away from any grading system used in your institution. Use a scoring system which is based on sound scoring rules and functions, and yields data which can be manipulated in well-defined ways, which make sense. That also means: going to the basics of what you want and try to do with your assessment. Take responsibility and initiative, don't trust so much (grading) systems offered by administration or higher levels in an institution or organisation, because these are by nature built with a distinct goal in mind. If you have a sound scoring system of your own, you can always find ways to transform or converse the resulting scores into grades that faculty or administration expects.

Remember: grades are not numbers, they are at most numerals (i.e. number symbols), or just alphanumeric symbols, or even graphical symbols like smileys. At most they can be ordered, but the distances between grades are not defined at all, numerically. Whatever the grades are or represent, they should be considered non-manipulable once you've got them. The real work so to speak should and will be done with scores, which are numbers obeying strict rules of operation like combining and refining, etc.

The university where I worked, the exam evaluation committee does expect a bell curve. It means they expect certain amount of students to be in all grades, from A to F. this actually has advantages and disadvantages. Like majority of the students fall in the B grade. A bell curve at times it is helpful when there is large number of students registered for a course, but it becomes disadvantageous if the cohort or batch is very small. In my opinion, The students may bring about competitive efforts if their grades are placed on a bell curve, for the forthcoming exams. The instructor may find it illogical sometimes.

One of the paradoxical consequences of using BCS (Bell Curve Smoothing) is the following (please, don't try it out, for the sake of your students).

Suppose you have to prepare an examination/test at the end of a course. Suppose you decide to only ask questions which you are quite sure that almost every student in your class will be able to answer completely and correctly. Indeed you find out that the vast majority of scores or grades fall at the two highest quality levels (small errors will still be made, for various reasons). Actually, all students would pass the test. But now you apply BCS, and suddenly 10% of all students don't pass the test!  WHY !?

The problem is evident: with BCS you are no longer testing / measuring cognitive and other competences, and whether students attain a certain minimal level of such competences, but you are basically putting students into a small number of ordered categories irrespective of the difficulty of the test task. And there will always be loosers, even if how they performed on the test would be considered OK by any common standard.

Of course, you can also construct the converse paradox with an even more drastic result: design a test so difficult that no student will be able to answer your questions completely and correctly, so that by all usual standards all but one or two students will pass the test.. Nevertheless, using BCS you suddenly will have miraculously 90%, i.e. 27 of 30 students, passing the test! How on earth is this possible !?

So it is test design that should be improved beforehand, not playing a dirty trick like BCS afterwards.

There is only one thing that let BCS stand out positively: it is cheaper than any serious test design, and it makes administration happy because they can now ascertain / certify that they have a constant throughput of 90%, irrespective of course content and irrespective of examination content. Fabulous!

Today I was reading again in the famous best-seller of Stephen R. Covey "The 7 Habits of Highly Effectove People", about Habit 4: Think Win/Win. Quite unexpectedly I came across a text in which Covey desrcibes in his own admirable way the intrinsic (but often hidden) problematic of using at all a Bell Curve, i.e. the normal distribution curve, in the context of educational judgement. Maybe this perspective on the question discussed here is even more telling and relevant than any criticism on Bell Curve use based on measurement considerations or the occurence of paradoxes.