This is related to "Skimming through a math paper with a group" and "What do professors gain out of teaching reading courses with individual Ph.D students", but from the other side :)

I've experimented with various ways to run my advanced Ph.D seminar, ranging from almost-lectures to "students present papers" to "students present textbook-level material" to "let's all work on a problem together". I don't think any of them have really worked to my satisfaction in the sense of ending the semester feeling that students have a command (as opposed to knowledge) of the material.

I've been reading about the Oxford tutorial style approach, which can crudely be approximated by:

- professor assigns reading material once a week
- students form pairs and meet with professor once a week for about 1-1.5 hours.
- students run the meeting (maybe one person presents and the other critiques, or they work out shared portions of the material on the board). Professor keeps quiet as far as possible except to unblock.

This format sounds tempting, as something that might work with a small group (at most 10 people). Does anyone have experience with this format and would it be suitable for advanced material at the graduate level ?

While I'm hoping it doesn't matter too much for this discussion, the topics for the seminar would be in theoretical computer science.

## 1 Answer

Best proseminar:

Long list of papers and book chapters, about three per week, carefully chosen to be easy to read at first, then more wiley.

We all had to write a prĂ©cis on the same one of the papers each week. The other papers were related, usually contradicting each other about the same topic. At least everyone had read one of the papers so we could discuss.

I learned how to read papers that way.

At the end of the semester we all had to assign three papers to the class once and lead that discussion.