Open brando90 opened 7 years ago
Possibly the Caesar system, which was used in 6.005: https://github.com/uid/caesar-web
do u like it? do u think its good? it would also be nice if it helps us make sure the code reviews actually happen
I liked using it while taking 6.005. It seemed to work fairly well, and might also be part of the solution for #39, provided that reviewers are attentive.
we dont want too much time being spent on code reviews. We want students to review other peoples code but only a few question. At least one. Not sure about an upper bound, cuz there is a trade off between perfect questions and many questions. Many questions could be more important.
at some point ML should be more similar to human learning and only require a few examples per question but many across a topic
Yeah, I think Caesar allows for the breaking up of code into different task sizes/chunks, so it should be possible to adjust how much code we assign for reviews (and I think it speeds up the process of commenting/review).
To help keep up the quality of the questions and to have users/students learn from each others work, can we set up some kind of code review system for them?