A dilemma, and a request for input

by Cliopatria

Well, here's an ethical dilemma with which I'd like some help.

For the past couple of years, I've been quietly keeping track of the ethnic and gender breakdown of my students and the grades they receive. I do this informally, mind you, and up until now I've kept the results entirely to myself.

But I've noticed some trends, trends that may speak to my teaching style and unconscious prejudices as well as to the varied levels of preparedness of my students. But it's such an explosive issue, that I am not sure I should put my own data out there. I'm not worried for my job -- I have tenure, and proving a bias case against me would be near impossible. I've got data to back up all my grading decisions. But there's no question that while I consciously bend over backwards to grade fairly, some groups are more likely to receive As than others.

I'm aware that class and social background often has a racial or ethnic dimension. I'm aware of the suggestion, widely discussed in recent months, that young men of all races are often less well-prepared for college work. And my own grading patterns seem to back that up. I've discussed my grading trends with other faculty members, who report similar results. This helps me realize that if it is bias on my part (which I don't think it is), I'm hardly alone -- at least three of my colleagues report similar results from their students.

Here's my question:

Given that I have students who read this blog, is it a bad idea to disclose the data? While I think there's some potential for fruitful discussion on this issue, especially when it comes to thinking up solutions, I'm worried about the impact on my current and future students. I want each person who enters my class to be certain that he or she will be graded solely on his or her work, not on sex or race. (And of course, I have A, B, C, and F students from every background -- I am talking broad generalities rather than hard and fast rules.) Is it possible that I could do real harm -- emotional or legal -- by mentioning that certain groups are statistically more likely to earn As?

I'd like to hear some thoughts, and I'll talk about it with some colleagues before I go forward with a post about my findings. Right now, I'm leaning against putting the statistics out there, but I'm not firmly decided yet. Polite input is appreciated.

comments powered by Disqus

More Comments:

Irfan Khawaja - 8/4/2006

I would not post the data.

Your sample is too small to be statistically significant, so discussing it in statistical terms is of questionable purpose and will yield findings of questionable validity. Very little will be gained, in my view, from a discussion of your statistics.

On the down side, once the data are out there, whatever your intentions, their presence will distract your students, put unwarranted suspicions in their minds, and undermine your rapport with them.

Since the overarching purpose here is to improve your capacity to teach, the negatives outweigh the positives. So my verdict is: don't.

Incidentally, I've been teaching 11 years, and while I haven't tracked ethnicity in my grades, I've been aware of the general issue and been alert to it. My experience shows no great ethnic divide with respect to grades.

The real problem here is that grading criteria differ so greatly from person to person (and in some cases, even from semester to semester for the same person) that few meaningful claims about causation can be made with anyone's statistics. One more argument against making the grades public.

Irfan Khawaja - 8/4/2006

Just to clarify, in the last part of my post, I didn't mean to imply that your grading criteria are variable, but rather that since the point of discussing your data is comparative, it would only have a point if you could assume that others' criteria were stable. But since they aren't, you can't profitably compare your data with anyone else's; and since you can't, your statistics merely report what is happening in your classes. On the face of it, the latter is not sufficient information for a useful debate in a public forum.

Barry DeCicco - 1/11/2006

"Your sample is too small to be statistically significant, so discussing it in statistical terms is of questionable purpose and will yield findings of questionable validity. Very little will be gained, in my view, from a discussion of your statistics. "

Validity is, of course the big question.

I would like to find out how to declare differences in a data set of unkown size and composition to be not statistically significant. It would be a great help to me. Right now I'm using statistical software to do this.

Barry DeCicco - 1/11/2006

A few comments, from a statistician.

You should sit down with a statistician, or a quant-heavy educational researcher, and discuss this. A lot of the discussion will be brainstorming/poking around into various root causes.

You should then try to identify *why* these differences occur, and what you could/should do to rectify them.

John H. Lederer - 1/11/2006

When you grade papers/exams do you make them blind (i.e. remove or block the student's name)?

In my long ago experience teaching I did this and found it a considerable help to be comfortable with what I was doing.

Andre Mayer - 1/10/2006

One related point: If you suspect bias in your grading (as opposed to your teaching, or in recruitment) this might be checked by seeking a "second opinion" on a group of exams or papers.

Hugo Schwyzer - 1/10/2006

I agree with a lot of the reasons to be cautious, here, which is why I will indeed hold back for now. After all, what I've got (given that I'm not a trained statistician) is really just anecdotal, and it may still say more about me than anything else. Reason for restraint. Thanks, folks.

Oscar Chamberlain - 1/10/2006

Hugo

It strikes me that the next thing you might do--if you have not done so already--is discuss the data and its collection privately with a statistician. Even if you are personally good with statistics, I would still recommend it. It is possible that there are aspects of your methodology--and that of your colleagues--that are skewing the "raw" data in certain directions.

It is also possible that your data reflects subtleties in student recruitment and admissions policies. For example, are African American students on your campus as representative a cross section of the larger African American population as white students are of the white population?

In short, given your own trepidation, there's a lot to ask about your data before you publicize it.

Jonathan Dresner - 1/10/2006

I don't understand, offhand, why you can't discuss the issues without discussing the grade data. "XYZ students do not seem to respond to my teaching thusly" would do just fine, without you needing to specify the "distance from mean" and standard deviation of the grade distribution by which you came to that conclusion.

Alan Allport - 1/9/2006

And yes.

Robert KC Johnson - 1/9/2006

I think you'd have to talk with your college counsel's office. The Buckley amendment's wording wouldn't seem to prevent disclosing (and then talking about) these kinds of stats, but the amendment has often been applied in remarkably broad (and unanticipated) ways.