In part I, I began describing some of the nuts and bolts of implementing specs grading. In part II, I’ll finish describing some of the finer details of the system, and also discuss how MBT fits naturally into a specs system.
As Nilson describes in her book, the all-or-nothing nature of specs grading encourages students to do professional quality work the first time. On the occasion that a student turns in non-passing work, she suggests granting a handful of revision opportunities in the form of “tokens”. I have usually granted students five tokens per semester, to be cashed in for either a 24-hour extension on anything with a due date OR a free revision of non-passing work. This gives students an opportunity to learn from their mistakes (a key component of MBT as well), but not so many opportunities that you are swamped with grading. Only once have I ever had a student do more than one revision.
One of the challenges in administering a specs grading system in a math course is that it’s not always possible for a student to determine whether or not his/her work is correct. Think back to the days when you were first learning to write proofs; did you oversimplify the argument? Did you attempt to use a theorem that didn’t apply? These mistakes can be difficult to spot in one’s own work (even now, if we’re being honest!). Talbert’s elegant solution was to expand from the two-level pass/fail rubric to a three level pass/progressing/fail rubric. A student earned a ‘Progressing’ designation if their work meets all specifications but contains some significant flaw in reasoning/computation. Work assessed at progressing is given one free revision (subsequent revisions on the same work must use a token).
Assigning a final grade
In the end, of course, we need to assign a final grade. Nilson breaks them down into two main categories: more hurdles and higher hurdles. The “more hurdles” model rewards students who pass more learning objectives with a higher grade. The higher “hurdles model” rewards students who pass more complex objectives/assignments with a higher grade.
I’ve settled on a straightforward “more hurdles” model: I essentially count the number of outcomes students have passed and assign the grade which corresponds to that number, as outlined in part on this table (where “Initial” describes student progress on learning outcomes on homework and “Secondary” describes student progress on learning outcomes on exams; note that as of this writing, these numbers are not finalized and may be subject to change):
In this way, there are no statistical/numerical calculations (which confounds our learning management system’s attempt at computing a current grade) involved in the final grade computation. See the syllabus for a more straightforward description of the system.
Traditional MBT in a specs graded course
You may already be seeing how easily MBT fits into a specs graded course. In a traditional MBT course using in-class exams you identify a list of skills/objectives students should pass in order to pass the course before the semester begins. Then that number is fed into a weighted average which computes a final grade.
In a specs graded course using MBT, no formula is needed: just count the number of exam problems passed by the end of the semester, and use that number (along with the amount of passing work in any other categories) to compute the final grade. I have done this in several calculus courses, and will be doing it again in calculus and linear algebra this fall. Students find that the MBT philosophy of exams fits very naturally with a specs graded course. In short, it’s not enough to get 7/10 on everything to pass; you must do quality work.
As you can hopefully see, specs grading and MBT share a lot in common; an emphasis on doing quality work and revision as a key part of the learning process are just two. I’ve found specs grading to be much more pleasing to use than traditional points-based grading, though I’m not convinced it’s right for every class. If you have any questions, feel free to sound off in the comments or send me an email.