Initial Publication Date: August 16, 2018
Students are not unit tests: providing assessment that is fair across students with different types of skills
Benjamin Bratton, Molecular Biology, Princeton UniversityMy main instructional responsibility is in helping teach a course called "Integrated Science Curriculum" a transdisciplinary approach to teaching introductory science. The students receive instruction from the departments of computer science, physics, chemistry and molecular biology in a single, integrated course. Altogether, there are roughly 15 faculty and staff who are responsible for this teaching load and they are drawn from their respective departments into a cohesive team. This means that the students are necessarily looking for connections between the disciplines and using the tools and language of one discipline to describe another. Summative assessment is in general difficult with students that have skills in different disciplines and there is an additional structural aspect that adds complexity to our situation.
The first difficulty that I have experienced is about the nature of computational questions. To borrow an image from computer science, unit tests either pass or fail. What I mean is that if the student is using a computational approach to answer a question, they can only get the correct solution if their approach and implementation are both correct. Imagine a student A who is able to see their way to a solution clearly but is not able to implement it and has many difficulties with syntax. On the other hand, student B is able to find brilliant ways to implement a method but struggles with knowing how to connect that with the biological or physical context of the question. Full success requires both of these skills. We could offer credit only for a fully correct solution, but it seems as if options for partial credit (non-boolean response) should be available. But if student A receives partial credit for a good setup, how can B receive partial credit for execution of a non-solution? While we haven't been incredibly successful at combating this issue, we've tried two ideas with some success. The first is to recognize that interdisciplinary science rarely happens in a solo context, so we encourage our students to work on their homework sets together. We even provide paid tutors from previous cohorts who come to the weekly problem session to help give hints and tips. The second is to provide questions that explicitly test the latter type of skills by providing skeleton code or code that doesn't run and ask the students to edit the code so that it completes the desired task.
The second, institution specific, issue that we have is our institution prohibits the use of computers/phones/computational aids other than a scientific, non-graphing calculator during in-class assessment. In general, this means that we choose to assess the student's ability to set-up a problem or an approach and limit the amount of numerical execution required. However, this approach has left us in the somewhat unsatisfying situation where we assess computational skills in a separate exam. While this may not be the ideal solution, we also have weekly problem sets and lab reports and include a mix of fully numeric computation problems as well as questions that simply require setup.