On Assessment, IV

Last week, I had ended with an interpretation for an A–D grading scale, shown below (here is a link to last week’s post for reference).

Day166grades

I remind you that this scale is not ideal; the purpose was to come up with some system of assigning grades which wasn’t punitive, but rather which motivated students to learn concepts rather than to avoid losing points on exams.

We continue with a discussion of how to use such a system in practice.

“Now let’s consider this in the context of an exam.  The first part of an exam is a skills portion, with, say, ten short problems of roughly equal length.  Expectations for this part of the exam [meaning a grade of B] are seven problems “essentially” correct, and four problems completely correct.  These expectations are written on the exam for students to see.

Are these expectations too low?  Perhaps.  But then an A [refer to the chart below] means eight problems “essentially” correct, with at least five completely correct. Of course we must ask what it means for a problem to be “essentially” correct — but when in doubt, err on behalf of the student.  (Students rarely suggest that their scores be lowered.)

Then grading is actually somewhat easier, and grades can be assigned as follows, with the abbreviations EC and CC meaning essentially correct and completely correct, respectively (for simplicity, a grade of C is assigned for all other cases not accounted for):

Day167grades

Now this eliminates the need for partial credit — but does require a judgment as to what “essentially correct” means.

This also makes grading much easier.  I would suggest that each problem be marked as “EC,” “CC,” or left blank.  Few comments, if any are necessary.  This is the approach I have taken in BC Fast-Track, and it encourages further learning as it leaves the student in the position of needing to work through their mistakes.

I would have students keep a section of their notebooks for exams and revisions, and there they can keep their reworked problems, should they choose to do so.  Then — as I did in BC Fast-Track — students could visit me periodically with their notebooks and I can take a look at their ongoing progress.  This “additional” work, if sufficiently well done, could boost their grade at the end of the semester.

I think this could have the same effect it did in BC Fast-Track — exams were easier and more enjoyable to grade.  But there were more discussions in my office about reworked exams and sources of error that were initiated by the students themselves, and these discussions were not about points, but about concepts.

Now what about the part of the exam which is intended to be more conceptual?  Let us suppose that there are three problems, roughly comparable in length, and of various difficulties.  Then grades might be assigned as follows:

Day167grades2

More details about how this would fit in a classroom environment may be found in a later document [I cannot recall which document is being referred to here].  But this system allows for a more qualitative approach to grading.  Performance expectations are also clearer, but such expectations depend critically upon the nature of the problems given.  Moreover, grades are not assigned punitively, but the emphasis is on doing problems completely and correctly.

For an example, below could be a set of ten skills problems and three conceptual questions for a basic assessment on the rules of differentiation.  This would a 70-minute assessment.  Given expectations for completely correct problems, I think this is reasonable.

Skills questions:

1.  Evaluate

\lim_{x\to\pi/4}\dfrac{\tan x-1}{x-\pi/4}.

2.  If f(x)=e^x\cos x, find f^\prime(x).

3.  Find

\dfrac d{dx}\dfrac 7{\sqrt{x^3}}.

4.  Using the quotient rule, find

\dfrac d{dx}\dfrac{x^3}{\sin x},

simplifying as much as possible.

5.  Find the derivative of f(x)=\left(\sin\sqrt x\right)^{\!2}.

6.  Find the equation of the tangent line to h(x)=\sec(2x) at x=\pi/6.

7.  Using a definition of the derivative, find the derivative of p(x)=x^2-x.

8.  Assume that f and g are differentiable functions. Find

\dfrac d{dx}f(g(x^2)).

9.  Find the derivative of q(x)=x^2e^x\cot x.

10.  Let f be the greatest integer function. Using the definition of the derivative, determine whether or not the derivative exists at x=0.

Conceptual questions:

1.  Using the product rule, find

\dfrac d{dx}f(x)(g(x))^{-1}.

Explain your result.

2.  Suppose that the line y=6x+a is tangent to both f(x)=x^2+b and g(x)=x^3+3x. Find a and b.

3.  Suppose that f is a differentiable function. Discuss the following limit:

\lim_{h\to0}\dfrac{f(x+2h)-f(x-h)}{3h}.

Stay tuned for next week’s final installment of this series on assessment….

On Assessment, III

Today, I’ll continue with the discussion of assessment I began a few weeks ago.  Last week, I ended with the observation that “The performance of our students determines our expectations, rather than the other way around.”  Before continuing, I just want to make clear the context for that remark.  When I wrote this essay, IMSA was fully funded by the state of Illinois.

What would it look like to legislators — who approved our funding — if you brought in the top students (relative to their local regions), and then they regularly received low grades, or worse yet, flunked out?  Not good.  And because the admissions process was fairly involved, would this indicate a major flaw in this process?

Yes, some students couldn’t handle the high-pressure academic environment.  But I know that in several of my classes, a student needed to work hard not to earn at least a C.  In other words, showing up, handing in homework on time, and doing a reasonable amount of studying for quizzes and exams would usually guarantee a grade of at least a C.

Now is not the time to dive more deeply into this artificial adjustment of expectations, but I did want to mention that this issue is a significant one, even at presumably elite schools for mathematics and science students.

So to continue with the essay….

“Now there is a natural give-and-take between evaluating student performance and setting expectations.  And, of course, the above remarks are nothing but generalizations.  But they illustrate some of the important issues at hand, and may bear fruitful discussion.

Moving to more concrete issues, I believe that the assignment of letter grades on exams in BC Fast-Track [recall, this was the colloquial name for the Honors Calculus sequence] was, on the whole, successful.  Without going into unnecessary detail, the classroom environment was such that the assigned grades were meaningful to the students.  To give a few examples, I assigned a grade of A+ for truly outstanding work, perhaps only a half-dozen times throughout the entire semester.  The students knew this, and so that accolade truly meant something.

Moreover, an A meant something as well.  It was truly rewarding to see the real pride of a student who, used to earning grades in the B range, began to earn the occasional, or perhaps more frequent, A.  Admittedly, students who made it to the second semester were essentially guaranteed a grade of no lower than a B-.  But this seemed to make an A that much more meaningful.

So student exams had two letter grades on them — one for the skills portion of the exam, and one for the conceptual portion of the exam.  No points were assigned, and few comments were made.  Students were expected to rework problems on which they made errors.

I bring up this point because I think this system of assigning grades really did motivate students to learn calculus rather than accumulate points.  This is the critical issue: I suggest that the way we assign grades does little to disabuse many students that taking a mathematics course is about accumulating sufficiently many — or losing sufficiently few — points.

Let’s take a particular example.  The past few semesters, I stopped assigning half-points on assessments [the usual practice at IMSA].  I might forgive a sign error now and then, but too many on a single assessment would warrant a point or two off.

In the past, I simply considered a sign error as a half-point off.  And so it was.  But consider that without being able to occasionally perform fairly involved calculations, it is not possible to become a successful mathematician.  Attention to detail is as important in mathematics as it is in any number of other disciplines, and we try to develop that skill punitively — you don’t attend to detail, and we will take off points.

Of course one might argue that points are given for work well done — but any of us could, I think, agree that when discussing the grading of an exam, it’s how many points off for a particular type of error that is discussed as much as, or even more than, how many points are given for work correctly done.

And so the idea of “partial credit” is born.  Perhaps now is not the place to begin this discussion, but consider that a student might meet expectations (that is, earn a B) without ever having done an entire complex problem on an assessment completely correctly.  (Some teachers have even gone so far as to give no partial credit.  See On Partial Credit, Letter to the Editor, MAA Focus, February 2002, p. 17.)

Why this system of points and partial credit?  One may speculate as to its origins, and there is controversy even now about its use on standardized exams.  But I cannot help feeling that one function of partial credit is that it allows a teacher to defend the assignment of a particular grade.  “Every sign error is a half-point off.  That’s why you got a B+ instead of an A-.   I have to use the exact same scale for everyone in order to be fair.

But doesn’t this simply shift the responsibility for the grade onto a rubric?  I suggest that many of us would feel competent to take a set of calculus exams — with names removed — and within five or ten minutes, separate out all the A papers.  Of course this is subjective — but no less subjective than saying that this problem is worth six points while another is worth ten, or that sign errors are a half-point off, unless, of course, the derivative of the cosine is taken incorrectly, in which case it’s a whole point.

Thus the assigning of points is no more “objective” than giving a letter grade.  As I’m sure that anyone who has graded a complex word problem based on an assignment of points can attest to.  Consider the student who has the entire procedure correct, but because of a few algebra errors, has no intermediate calculation correct.  The problem is worth ten points.  How many points should the student receive?

Well, of course, you say you’d have to see the problem first.  But I say, no.  The student receives a C.  Having no intermediate calculations correct demonstrates — regardless of what else — that the student has not met expectations.

So is it possible to avoid points altogether?  Perhaps.  Consider the following grading system:

Day166grades.png

The essay continues in the next installment of On Assessment with a discussion of how to implement this system in practice.