Course grade control

CSGnet Archive CSG2001
1

[From Bruce Abbott (2001.05.05.0955 EST)]

Dick Robertson, 2001.0502.1735 CDT to Chuck Tucker --

I think you would have to
do something like I did, that is to have a student repeat his desired course
grade before taking each test, have him aware of his running average, score his
test immediately after taking it, have him recompute his running average and
subtract that from his desired grade ( I.e. say his desired grade is A {=.90 -
1.00, theoretically}, and his running average is now now .87, then his RS

-PS, or

error is -.03. In a true control situation he show study harder and retake

this

test, or do enough bettter on the next test to get his running average >.90.

Let's think about what this model implies about student behavior with
respect to the student's course grade. First, it says that students work to
control their course grade at some level (not necessarily the "A" level).
Given my experience, that seems reasonable to me. I suspect that all
students would "like" to receive an A for the course, but that many would
"settle" for something lower. On our campus as on many others, students
taking some courses (outside their major) can opt for a Pass/Fail grading
scale, where Pass is C or better. For many of those students there is
little incentive for doing more work than will yield a C. Others my be
taking the course under the usual letter-grade system but would be quite
happy just to pass the course.

Secondly, the model says that students keep track of how they are doing in
the course (i.e., they maintain a perception of their current course
performance). Most students I know do demand, receive, and record the
scores they receive on exams and papers and other numbers (e.g., completed
assignments) that inform them of their progress and performance in the
course. Many also like to know how well they have performed on a given test
or assignment _relative_ to other members of the class. This information
may be used to alter grade reference levels -- if hardly anyone in the class
gets an A, then perhaps the test or assignment was "too hard" or the
instructor has unreasonable expections about student performance. The
student may continue to "want" an A but realistically "expect" only a B.
How does this affect the student's reference level? The model doesn't say.

Thirdly, the model says that students compare the current perception of
course achievement to their reference levels for same, creating an error
signal. In PCT, error signals are not perceived, so a somewhat different
model is required to represent error that students do perceive between what
they wish for in the way of a course grade and what they perceive they are
likely to actually receive. However, let's set that aside as this
difference in detail does not affect the model's workings.

Fourthly, the model says that when there is error between reference grade
and perceived grade, students will take action to reduce the error. Is this
a one-way or a two-way system? That is, will students vary their actions
only to correct shortfalls or will they also vary them to correct
perceptions that the grade is rising above reference? That question leads
to the following one, which is perhaps more interesting than it may at first
seem: Why would a student take action to _lower_ a grade?

The model says that such action would be taken because the grade has risen
above the student's reference grade. In effect, the student is saying "I'm
doing better than I care to do, so I can afford to 'back off' on my efforts
here." Not visible in this model are the reasons why the student would want
to reduce effort. There are many possible reasons, but all of them come
down, I think, to conflict between control systems. There are many demands
on a student's time, and the student may be able to better control some
other perceptions by reducing the time spent studying or doing assignments
for the course in question. Or, studying may be perceived as effortful and
unpleasant; the student can bring the levels of these variables closer to a
zero reference level by reducing the time spent on studying, so long as this
does not bring course performance below the student's reference level.

But the model also says that a student can reduce error in the grade control
system by resetting the grade reference level so that it is closer to the
current level. When will this happen? The model says this will happen if
there is "persistent error" in "essential" variables, thus increasing the
rate of reorganization. Does persistent error in one's perceived grade
affect a person's essential variables? If so, which ones, and how? The
model doesn't say.

The fact that course grade is being set up as the controlled variable in
this model leads to a difficulty that I have not seen mentioned. Because
course grade is based on _cumulative_ performance, there comes a point at
which earlier shortfalls in performance cannot be corrected for by any
amount of increased effort. If the student has failed the first two
quizzes, for example, an A may be out of the question no matter what the
student does after that. The student will usually know this, of course.
How will the student's reference be affected? Will the student operate in a
region of large error for the remainder of the course? This would lead to
an increased level of effort aimed at doing the best possible work under the
circumstances. Will the student reset the reference level to something
attainable? Will the student give up attempting to control the course grade
(e.g., by dropping out)? The model doesn't say.

These considerations suggest that an elementary control-system model of
student course grade control is not going to be adequate. That model does
not specify how the reference should change and does not include the effects
of conflicting control systems, limitations on output resources, the
constraint on course grade imposed by earlier course performance, etc. In
short, it's too simple and as it stands, can account for virtually any
outcome post hoc.

I'm not saying that this effort should be abandoned, but I do think that
some careful thinking needs to be done _in advance of collecting data_ to
develop a more adequate model or models. The model(s) will then tell you
what additional variables you need to collect data on in the course of the
investigation.

Bruce A.

2

[From Dick Robertson, 2001.0509.0955CDT]

Abbott_Bruce wrote:

[From Bruce Abbott (2001.05.05.0955 EST)]

Let's think about what this model implies about student behavior with
respect to the student's course grade. First, it says that students work to
control their course grade at some level (not necessarily the "A" level).
Given my experience, that seems reasonable to me. I suspect that all
students would "like" to receive an A for the course, but that many would
"settle" for something lower. On our campus as on many others, students
taking some courses (outside their major) can opt for a Pass/Fail grading
scale, where Pass is C or better. For many of those students there is
little incentive for doing more work than will yield a C. Others my be
taking the course under the usual letter-grade system but would be quite
happy just to pass the course.

Bruce,

I like your thoroughgoing analysis of the grade control study. Your point about the

errr signal not being percetable in PC theory was one I had'nt thought of when
I did the study. It is worth considering.

Fourthly, the model says that when there is error between reference grade
and perceived grade, students will take action to reduce the error. Is this
a one-way or a two-way system? That is, will students vary their actions
only to correct shortfalls or will they also vary them to correct
perceptions that the grade is rising above reference? That question leads
to the following one, which is perhaps more interesting than it may at first
seem: Why would a student take action to _lower_ a grade?

According to some comments I seem to remember the answer was Yes, for
some people.

The model says that such action would be taken because the grade has risen
above the student's reference grade. In effect, the student is saying "I'm
doing better than I care to do, so I can afford to 'back off' on my efforts
here." Not visible in this model are the reasons why the student would want
to reduce effort. There are many possible reasons, but all of them come
down, I think, to conflict between control systems. There are many demands
on a student's time, and the student may be able to better control some
other perceptions by reducing the time spent studying or doing assignments
for the course in question. Or, studying may be perceived as effortful and
unpleasant; the student can bring the levels of these variables closer to a
zero reference level by reducing the time spent on studying, so long as this
does not bring course performance below the student's reference level.

But the model also says that a student can reduce error in the grade control
system by resetting the grade reference level so that it is closer to the
current level. When will this happen? The model says this will happen if
there is "persistent error" in "essential" variables, thus increasing the
rate of reorganization. Does persistent error in one's perceived grade
affect a person's essential variables? If so, which ones, and how? The
model doesn't say.

The fact that course grade is being set up as the controlled variable in
this model leads to a difficulty that I have not seen mentioned. Because
course grade is based on _cumulative_ performance, there comes a point at
which earlier shortfalls in performance cannot be corrected for by any
amount of increased effort. If the student has failed the first two
quizzes, for example, an A may be out of the question no matter what the
student does after that.

Not quite. Each student had to pass the test for each chapter before going on to
the test for the next chapter. He or she could pass it with 60% and move on, and
come back later to pass it at a higher level. They could retest the same chapter as

many times as needed to get to the minimum level.

The student will usually know this, of course.
How will the student's reference be affected? Will the student operate in a
region of large error for the remainder of the course? This would lead to
an increased level of effort aimed at doing the best possible work under the
circumstances. Will the student reset the reference level to something
attainable? Will the student give up attempting to control the course grade
(e.g., by dropping out)? The model doesn't say.

These considerations suggest that an elementary control-system model of
student course grade control is not going to be adequate. That model does
not specify how the reference should change and does not include the effects
of conflicting control systems, limitations on output resources, the
constraint on course grade imposed by earlier course performance, etc. In
short, it's too simple and as it stands, can account for virtually any
outcome post hoc.

I'm not saying that this effort should be abandoned, but I do think that
some careful thinking needs to be done _in advance of collecting data_ to
develop a more adequate model or models. The model(s) will then tell you
what additional variables you need to collect data on in the course of the
investigation.

Yeah, I'd love to see someone who is still teaching go on with a refined model.

···

Best, Dick R.

3

[Chuck Tucker (2001.05.20.00EDT)]

In a message dated 5/5/2001 11:06:29 AM Eastern Daylight Time,
ABBOTT@IPFW.EDU writes:

<< [From Bruce Abbott (2001.05.05.0955 EST)]

>Dick Robertson, 2001.0502.1735 CDT to Chuck Tucker --

>I think you would have to
>do something like I did, that is to have a student repeat his desired course
>grade before taking each test, have him aware of his running average, score
his
>test immediately after taking it, have him recompute his running average and
>subtract that from his desired grade ( I.e. say his desired grade is A
{=.90 -
>1.00, theoretically}, and his running average is now now .87, then his RS
-PS, or
>error is -.03. In a true control situation he show study harder and retake
this
>test, or do enough better on the next test to get his running average >.90.

<<Let's think about what this model implies about student behavior with
respect to the student's course grade. First, it says that students work to
control their course grade at some level (not necessarily the "A" level).
Given my experience, that seems reasonable to me. I suspect that all
students would "like" to receive an A for the course, but that many would
"settle" for something lower. On our campus as on many others, students
taking some courses (outside their major) can opt for a Pass/Fail grading
scale, where Pass is C or better. For many of those students there is
little incentive for doing more work than will yield a C. Others my be
taking the course under the usual letter-grade system but would be quite
happy just to pass the course.>>

This brings up the issue of the type and "structure" of the course that is
being taught. Most of the students (about 90%) in my Intro Socy (101) were
NOT Socy majors and thus did not have to get at least a "C" in the course to
continue in the major. As is the case for all of the 101 courses taught each
semester in the department most of the students are taking the course to
fulfill a social science requirement for their own majors. In our grading
system a non-socy major can receive a "D" in 101 and still get credit for it.

I also offer a course where the class composition and grade requirements are
quite different. It is a Jr-Sr level course wherein over 90% of the students
are Socy majors some of which are hoping to graduate after the semester in
which I am offering the course. All Socy majors must obtain at least a "C"
in the course to be able to graduate and if they don't do this they must
retake the course. This course is required (no exceptions or substitutions)
to graduate with a major in sociology. I suspect that this course is more
demanding in "grade control" than other courses offered in the department.

<<Secondly, the model says that students keep track of how they are doing in
the course (i.e., they maintain a perception of their current course
performance). Most students I know do demand, receive, and record the
scores they receive on exams and papers and other numbers (e.g., completed
assignments) that inform them of their progress and performance in the
course. Many also like to know how well they have performed on a given test
or assignment _relative_ to other members of the class. This information
may be used to alter grade reference levels -- if hardly anyone in the class
gets an A, then perhaps the test or assignment was "too hard" or the
instructor has unreasonable exceptions about student performance. The
student may continue to "want" an A but realistically "expect" only a B.
How does this affect the student's reference level? The model doesn't say.>>

This brings up the importance of the grading procedures for a course.
Testing for grade control requires that the grading procedures be as simple
and straightforward as possible so the student will know after ever quiz,
test or paper what grade she has relative to her reference grade. I never
"curve" my grades so the students never are concerned about their grades
relative to others (this procedure BTW is a way of letting students grade
each other rather that an instructor). I use a point system which is easily
converted to grades so the student knows at any point in the course what her
grade is and what she has to do on the next test to maintain that grade.
There is no retaking of tests or "extra-credit" so the student is "stuck"
with the grade.

<< Thirdly, the model says that students compare the current perception of
course achievement to their reference levels for same, creating an error
signal. In PCT, error signals are not perceived, so a somewhat different
model is required to represent error that students do perceive between what
they wish for in the way of a course grade and what they perceive they are
likely to actually receive. However, let's set that aside as this
difference in detail does not affect the model's workings.>>

The student has to be able to perceive error in order adjust their behavior
either by "studying more" or by changing the reference grade.

<<Fourthly, the model says that when there is error between reference grade
and perceived grade, students will take action to reduce the error. Is this
a one-way or a two-way system? That is, will students vary their actions
only to correct shortfalls or will they also vary them to correct
perceptions that the grade is rising above reference? That question leads
to the following one, which is perhaps more interesting than it may at first
seem: Why would a student take action to _lower_ a grade?>>

This is a negative feedback model so it is a one-way system; there is only
correction for error. If the student's reference grade is "A" then it is
either success or an error. If the student's reference grade is "B" and she
obtains an "A" there is no error thus she has "over-controlled." I have
discovered that student's who do this will adjust their reference grade (to
an "A") or just maintain their study routine so they can obtain their
previous selected reference grade. A danger here is that a student might
define the course as an "easy" or "crip" course and reduce studying and
receive a lower grade on the next exam.

<<The model says that such action would be taken because the grade has risen
above the student's reference grade. In effect, the student is saying "I'm
doing better than I care to do, so I can afford to 'back off' on my efforts
here." Not visible in this model are the reasons why the student would want
to reduce effort. There are many possible reasons, but all of them come
down, I think, to conflict between control systems. There are many demands
on a student's time, and the student may be able to better control some
other perceptions by reducing the time spent studying or doing assignments
for the course in question. Or, studying may be perceived as effortful and
unpleasant; the student can bring the levels of these variables closer to a
zero reference level by reducing the time spent on studying, so long as this
does not bring course performance below the student's reference level.>>

This does happen. I find out about it when a student has lowered their effort
on the basis that my course is easy and then they receive a lower grade. They
come to me with the complaint that my course is deceptively difficult when
they are the one who made the decision to reduce their effort since their
grade was higher that expected. I try to explain this to them but it rarely
satisfies them.

<< But the model also says that a student can reduce error in the grade
control
system by resetting the grade reference level so that it is closer to the
current level. When will this happen? The model says this will happen if
there is "persistent error" in "essential" variables, thus increasing the
rate of reorganization. Does persistent error in one's perceived grade
affect a person's essential variables? If so, which ones, and how? The
model doesn't say.>>

In my grading procedures it happens when the student is convinced that it is
impossible to obtain the grade she "wants" and/or "expects." In my system if
a student has "lost" 11 points then the highest possible score they can get
is "89." An "89" is a grade of "B+" so if your reference grade an "A" then
you can't obtain it with an "89." You have to either adjust or live until the
end of the semester with this "error."

If by "essential variables" you mean "basic biological functioning" (e.g.,
blood pressure) I believe that these are only effected temporarily by grades
(although I have noticed that just prior to my final examinations a large
number of students claim that their relatives have died and they can't
possibly take my examination. Of course, I require a funeral announcement for
such claims).

<<The fact that course grade is being set up as the controlled variable in
this model leads to a difficulty that I have not seen mentioned. Because
course grade is based on _cumulative_ performance, there comes a point at
which earlier shortfalls in performance cannot be corrected for by any
amount of increased effort. If the student has failed the first two
quizzes, for example, an A may be out of the question no matter what the
student does after that. The student will usually know this, of course.
How will the student's reference be affected? Will the student operate in a
region of large error for the remainder of the course? This would lead to
an increased level of effort aimed at doing the best possible work under the
circumstances. Will the student reset the reference level to something
attainable? Will the student give up attempting to control the course grade
(e.g., by dropping out)? The model doesn't say.>>

Those students who are sitting in the class with a "deceitful or hateful
look" on their face may be those who know that there is no way they can
obtain their reference grade. It is your fault as the teacher (who "gives"
the grades) that they will not get their reference grade.

I have known some students who have "dropped the course" rather that receive
a lower grade (I advised my daughter to do this in a non-major course so she
would not lower her GPA and thereby effect her scholarship) but I have known
some students who refuse to drop the course for other reasons even though
they would receive an "F." Again, the "structure" of the grading system is
taken into account: by dropping the course they would lose their scholarship
since they were not enrolled in 12 hours of credit for the semester (this
effects student-athletes since NCAA requires tat they are enrolled in 12
hours of credit each semester). So, other matters may influence a student
when "controlling" for a grade.

<<These considerations suggest that an elementary control-system model of
student course grade control is not going to be adequate. That model does
not specify how the reference should change and does not include the effects
of conflicting control systems, limitations on output resources, the
constraint on course grade imposed by earlier course performance, etc. In
short, it's too simple and as it stands, can account for virtually any
outcome post hoc.>>

<<I'm not saying that this effort should be abandoned, but I do think that
some careful thinking needs to be done _in advance of collecting data_ to
develop a more adequate model or models. The model(s) will then tell you
what additional variables you need to collect data on in the course of the
investigation. >>

I would agree. The other matters that I have mentioned (and more) should be
taken into account in any study of grade control.

Regards,
              Chuck

4

[Chuck Tucker (2001.05.13.14:18 EDT)]

In a message dated 5/5/2001 11:06:29 AM Eastern Daylight Time,
ABBOTT@IPFW.EDU writes:

<< I'm not saying that this effort should be abandoned, but I do think that
some careful thinking needs to be done _in advance of collecting data_ to
develop a more adequate model or models. The model(s) will then tell you
what additional variables you need to collect data on in the course of the
investigation.

Bruce A. >>

A reconsideration of my (2001.05.20:00EDT) post leads my to say that it is
not possible to "test" for student grade control with my grading procedures
or most that I know about because they do not allow (as Dick's grading
procedures do) to correct for error. The grading procedures must allow for
the student to correct for an error at any time during the semester and if
she does not obtain the reference grade then allow for dropping of the course
w/o penalty. The only grading system that I know of which allows for dropping
a course no less that 24 hours before the final examination is at Stanford
University. I suppose that an instructor could make such an arrangement for
her course by back dating the drop form.

A colleague of mine uses a set of procedures wherein each student selects the
grade she "wants" in the course and is allowed to "redo" or "makeup" any
assignment at any time during the semester to obtain the grade. Very few of
his students fail to achieve their reference grade and, as I recall, those
who don't are usually dropping out of school completely.

Regards,
                Chuck

5

[Chuck Tucker (2001.05.14.12:00EDT)]

I checked with my colleague to see if my description of his grading
procedures were correct; they were not correct. Here they are:

I set standards for what constitutes a "good-faith" effort that when students
meet it, then I record what grade they tell me to record. If they do not meet
it, then I have set standards for B through D. If a lesser standard is met,
then no grade lower than that grade for that standard will be awarded.
Students do not have the opportunity to make up work at any time in the
semester. Homework (class preparation) is expected to be done for class. Unit
assignments can be retried but within a specified period. I discuss these
procedures and standards with the students and invite them to make
suggestions for improvements. They typicaly don't. Most students do receive
an A.

These procedures can't be used to "test" for grade control under the PCT
model but they do come close and could be modified slightly to fit the
requirements of PCT.

Regards,
                Chuck