Christian Kästner has done away with all in-person sessions, opting instead for meetings with teaching assistants, assessments and video demonstrations — whilst acknowledging that this deviates from best teaching practice.
Christian Kästner teaches an advanced course at Carnegie Mellon dedicated to the production of machine learning systems, which is taken by one hundred to one hundred and seventy students each year, with a large part devoted to MLOps. Noticing that agents were now able to process all of his assignments, he reviewed his evaluation methods.
The principle is simple: no longer measure comprehension based on homework. Evaluation is now based on direct exchanges with teaching assistants, exams and filmed demonstrations. What students have to learn, on the other hand, has not changed.
The teacher acknowledges that some of these choices contravene the recommendations of educational research, and nevertheless assumes them. The course, moreover, largely authorizes the use of AI — without obligation to report it, and sometimes by encouraging it — with the exception of written and oral exams.
Read the full article on the original source
Go to the source →