Dvi Specs
Loren Swaggr
Specs, founded in 1987, is the world's leading provider of compound management services and supplier of research compounds to the Life Science industry. The compound management services are offered from our two main logistic centers in The Netherlands and Maryland, USA. In these warehouses, millions of compounds from our clients are stored under controlled environmental conditions and are processed using state-of-the-art weighing stations, automated liquid handlers and quality control devices. After processing, the samples are distributed to the end users on a daily basis all over the globe. Compound sourcing and procurement is a service that our clients use for analog searching and library enhancement. Our synthesis lab can help out with custom synthesis or contract research if compounds are not commercially available.
The Specs in-house 350.000+ screening compound repository consists of single synthesized, well-characterized and drug-like small molecules and has been built through global acquisition programs utilizing a network of more than 2,000 academic sources worldwide. These compounds are available for ordering online through www.specs.net. Pre-selected targeted or diverse libraries are available in various formats and library sizes. Our cheminformatic service can help with target specific selections for lead discovery and optimization programs and design of new chemical entities. Specs has a 30+ years proven track record in every aspect of compound management. Our combined services makes Specs uniquely qualified as a reliable outsource partner for compound libraries and logistics.
Two and a half years ago, I decided that the traditional system of grading student work --- based on assigning point values to that work and then determining course grades based on the point values --- was working against my goals as a teacher, and I decided to replace it with specifications grading. I had just learned about specs grading through Linda Nilson's book on the subject. This happened right at the end of Fall semester 2014, and I spent the entire Christmas break doing a crash-course redesign of my Winter 2015 classes to install specs grading in them.
I've used specs grading fifteen times since then: once in Cryptography and Privacy, once in Abstract Algebra 2, twice in Calculus 1, four times in Discrete Structures 1, and eight times in Discrete Structures 2. It's fair to say that my implementation has been battle-tested and has undergone a fair bit of evolution in that time. The first attempt in Winter 2015 was pretty rough, but very promising. Every semester since then, I made changes and updates to try to address issues that students and I noticed.
But it was only this last semester, the one that just concluded this week, where I felt that at every point during the semester --- from day 1 all the way through turning in course grades yesterday --- the specs grading system I had in place was working the way I wanted. It's still not 100% there, of course, but I think I have a blueprint of how to use specs grading moving forward[1] and of course, I want to share it with everyone.
In specifications grading, instead of using points to assess student work, the work is graded on a two-level rubric --- that is, some variation on Pass/Fail or Satisfactory/Unsatisfactory. Instructors craft a set of specifications or "specs" for assignments that define what Satisfactory work looks like. When the work is handed in, the instructor simply categorizes it as Satisfactory or Unsatisfactory depending on whether it meets the specs or doesn't. There are no points, so there is no partial credit. Instead, instructors give detailed feedback on student work, and specs grading includes giving students the opportunity to revise their work based on the feedback, and submit a revision as an attempt to meet specs.
Specs grading still uses an A/B/C/D/F course grade reporting approach, but the letter grades are earned differently. Rather than calculating complex weighted averages of points --- which you can't do because there are no points --- letter grades are earned by completing "bundles" of work which increase in size and scope as the letter grade being targeted goes higher. The idea is that students who want a "C" in the course have to do a certain amount of work that meets the specs; those wanting a "B" have to do everything the "C" people do, but more of it and of higher quality and/or difficulty level. Similarly the "A" students do everything the "B" students do plus even greater quantity and quality.
Done right, specs grading allows students choice and agency in how and when they are assessed; students are graded on what they can eventually show that they know, and they get to learn from mistakes and build upon failures; their grades are based on actual concrete evidence of learning; and the grades themselves convey actual meaning because they can be traced back to concrete evidence tied to detailed specifications of quality. The instructor often saves time too, because instead of determining how to allocate points (which takes more time than you think), she just determines whether the work is good enough or not, and gives feedback instead.
The specs grading setup I am going to describe here is for Discrete Structures 2, a junior-level mathematics course taken almost exclusively by Computer Science majors. It's the second semester of a year-long sequence and it focuses on mathematical proof and the theory of graphs, relations, and trees. I think that much of the structure I am going to describe here could be ported to other math classes, though.
Some people may debate whether or not "engagement" ought to be part of the grade. Personal experience with this course tells me that it should, in this case. What I mean here is not just attendance in class, but also preparation for class, active participation during class, and enagagement in the course outside of the class. I want students to treat the course as a high priority and engage with it as such.
If these are the things I want from the course, then I need to set up stuff for students to do and submit to me that will allow me to measure whether or not they are progressing or succeeding in those areas.
For basic technical skills, I combed through the course and decided on a list of 20 basic skills that I felt were essential building-block skills for the course. Those are called Learning Targets. These were keyed to the four major topics in the course (proof, graphs, relations, trees). Here are a couple:
For ability in application of these basic skills, students were given a series of Challenge Problems. These are problems that require students to apply what they learned about the basic skills and included a mix of "Theory" problems that involved writing proofs, programming assignments where students had to write Python code to solve a problem, and real-world applications. I started with a core of ten Challenge Problems but also wrote some more during the semester as I got inspired, and we ended up with 17 of these total. Here's one that involved doing some proofs by induction. Another had students write a Python function that would compute the composition of two relations on a finite set. Another had students use Python code to experiment with a class of graphs, make a conjecture about their clustering coefficients, and then prove their conjecture.
Finally, for engagement, I broke from the specs grading mold and used points, or what I called engagement credits. Students accumulated engagement credits through the course for doing things like completing their Guided Practice pre-class work on time and participating in class on certain days (especially the days close to breaks). Basically this was a way of incentivizing student work on the course for things that I needed them to do, especially outside of class.
The Learning Targets were assessed using short quizzes called Learning Target assessments. I set aside every other Friday in the course for students to take Learning Target assessments as well as a few extra days during the semester. Here's an example of the assessment for Learning Target P.2 and here's the one for G.6. Notice they are simple tasks that deal directly with the action verb in the Learning Target.
Students could come on these Fridays and take as many or as few of these Learning Target assessments as they wanted. Only the Learning Targets that we'd discussed in class were available, but once they were available they were always available. Previously-given Learning Target assessments would have new versions of the same problem available to do. So a student who didn't feel ready to be assessed on Learning Target G.6 didn't have to take the assessment for G.6, but just wait two weeks and try it then.
Learning Target assessments were graded Satisfactory/Unsatisfactory according to specifications that I determined, and those specs are at the bottom of each Learning Target assessment so it's very transparent for everyone. If student work was Satisfactory, I just circled the "S" at the top of the page, and circled "U" otherwise.
Challenge Problems were not graded Satisfactory/Unsatisfactory but rather using the EMRN rubric, which is a modification of the EMRF rubric I wrote about here.[2] "Pure" specs grading would say that I should not make things so complicated, and just use Satisfactory/Unsatisfactory with a high bar set for Satisfactory. But I've found that in math classes, written work is hard to get right and students get easily discouraged, so I felt there should be some detail added to the 2-level rubric to distinguish between Satisfactory work that is excellent versus merely "good enough", and Unsatisfactory work that is "getting there" versus that which has major shortcomings. Students would submit their Challenge Problems as PDFs or Jupyter notebooks on Blackboard; I'd grade it there and leave feedback, then students could revise (see below).
c80f0f1006