This quarter, our instructors communicated to students through their syllabi that in classes taught with many parallel lectures, overall grade determination will be made collectively by all of the course instructors. Our goal in this communication was to reinforce our long-held commitment to ensuring that the grade a student receives in a class is determined by the mastery of the course learning objectives that they demonstrate, and is not affected by which instructor they have or which lecture they are enrolled in. We provide this statement to explain the grading policy and its rationale.

By introducing this new policy, we ended a policy that some of our instructors have previously adopted of publishing grading thresholds (e.g., “90% and above  = A- or A”) in syllabi. Although publishing grading thresholds has the benefit of making transparent to students what grade they can expect to receive in a class, it has led to many other unfortunate outcomes. Most concretely, 90% is not a desirable threshold for success in a math exam. Indeed, experienced instructors typically write exams with median scores between 60% and 70%, and in our international peer math departments, typical median exam grades are even lower. These ‘low’ scores result from asking mathematically challenging questions that require students to synthesize knowledge, creatively solve problems, and make correct and well-defended arguments. These are formative skills for undergraduate students in all majors; they are the hallmarks of the UCLA undergraduate experience, and they remain in high demand by employers and graduate schools, particularly in an era where AI tools can perform many routine calculations. Deep understanding is prioritized. When instructors write their exams so that the median score meets a specific, predictable and high threshold (e.g., 80-90%), instructors are often forced to rely on predictable exercises, predominantly in the form of calculations that test students’ ability to apply formulas, but do not fully cover the course’s learning objectives, or equip the students for success in their future classes. Low scores do not translate into low letter grades because the instructor is not expecting a student to show perfect mastery of absolutely every topic covered in the exam. Moreover, exams with lower averages (and a wider spread of scores) are beneficial because minor mistakes have less impact on a student’s final grade. The instructor learns better what the student knows, rather than penalizing them for the parts that they don’t.

What will grading coordination entail? The department is working to produce grading guidelines for each class with parallel lectures. These will take the form of grade windows (e.g., suggested upper and lower bounds for the percentage of As and A-s, and similarly for B+s, Bs and B-s), that reflect both the recent history of the class (so that students who take the class this quarter will see similar grades to students taking the class last quarter or in the same quarter last year), as well as department recommendations based on statistical analysis of student mastery levels, measured against the course learning objectives, for typical presentations of the class. The existence of a grade window allows an instructor to exercise their judgement and should ensure that students do not feel themselves to be in competition for a fixed number of As or Bs. For example, if students in a class do very well, then an instructor will likely award a number of As and A-s that is near the upper bound of the grading window. As teaching leaders, we also remain committed to working with any instructor who finds the grading window doesn’t represent the actual mastery shown by their students, and exceptional grade distributions, though they are expected to be rare, are not forbidden.

Marcus Roper, Undergraduate Vice Chair 

Will Conley, Professor of Teaching

Michael Andrews, Director of the Program in Computing