By Joanna Goode
Thanks to the hard work of the ACM Education Policy Committee and other organizations working to strengthen national policy support for computer science, there have been many important policy victories for computing education at the federal level. Issues of teacher certification, professional development, and curriculum have been highlighted as major topics which need to be addressed and strengthened.
As a community, I think we also need to start thinking hard about assessment. How can we measure student learning in a computer science classroom? How can we quantify what students know and can do? This is a challenge for a variety of reasons:
* Grant agencies and local educational agencies often want to see test score improvements to rationalize the existence of computing courses. So far, they often suggest looking at whether standardized test scores in math and/or science increase as a result of taking a computing course. To me, this seems to be measuring the wrong content knowledge. We don’t assess geography knowledge by seeing how literacy scores raise, right?
* Traditionally, computer science courses rely on one programming language and assess learning through the writing of programs. But, as we move away from a programming-centric version of computer science towards a more comprehensive model, how do we assess the rich breadth of the field without relying on writing programs in a particular language?
* There has also been an emphasis on the creativity of computing. How do we measure creativity in computing on a standardized test in which there is typically only one “correct” answer?
In thinking about this problem, I have come up with two different approaches to solving our assessment problem in computer science education. First, I think that much like art, a portfolio approach might be a good measure to show students’ breadth of knowledge about computing, while also highlighting the creative solutions that we want students to derive as part of their learning. Second, I think we might want to develop some test-type items that are aligned with the items offered on the NAEP tests, which currently assess student learning in a variety of other subjects (arts, civics, economics, geography, mathematics, reading, science, US history, and writing). These could be given to students as pre-tests and post-tests when they enroll in computer science courses, to demonstrate whether or not they are developing computing knowledge and skills as a result of the course.
What other ideas do CSTA members have about assessment?
How do you assess learning in your class?
Joanna Goode
CSTA Board of Directors
Why not teach Computer Science as an integration of logic and discrete mathematics. Now you have something that is more academically tractable—by the rules that everyone understands and accepts. Programs are proofs, if crafted in that way. Comments are justifications, etc.
I allow my students to write in any language they choose. I grade their preconditions, postconditions, their descriptions of invariants and ultimately their understanding of the algorithm and relationship of its parts.
This narrow focus on a particular programming language is emblematic of just the kind of stove-piping that our critics love to use to marginalize computer science. But, lest we place too much of the blame on teachers, I think that ETS and popular culture has contributed greatly to this sad state of affairs.
Of course, all of this comes with a cost (or two); it’s pretty difficult to find students who will absorb this kind of content because they have been raised on a diet of 19th century mathematics and science for the last ten years of their academic lives, and many teachers will be uncomfortable with that content—for a variety of reasons. As for the first: that’s why they are in your class. Be prepared to teach that content alongside the “fun” stuff. Regarding the second statement, perhaps time and attrition will remedy that situation, or, and this is more likely, CS will vanish in the American Public School classroom.
Of course, your results may vary.
TomR