(If you don’t know this quote.)

Who should a defense talk be given for?

For the reviewers of your thesis (who have read the thesis in detail anyways)?

For the chair of the committee (you usually does not read your thesis in advance)?

For your colleagues (who will at least share some common background knowledge with you)?

Or for your family (you might share the last name with you but probably not a lot of knowledge about your research subject)?

Ultimately, it depends on your own goals and also on the committee members. Still, if somebody asked me for advice on this topic now I’d tell him (from my own experience) that the talk should not be “too accessible” (as was the case for me). At least in the last 10 minutes or so just loose the rest of the crowd, throw in some proofs, go into depth, show off a bit.

In the end, this issue is probably not too crucial though as the grade of your thesis (… in Germany you get a grade for your thesis …) is hardly ever affected by the presentation you give (at least in the domain of computer science). So I’m still glad that I got positive feedback from friends and colleagues outside the domain of information retrieval, even if the “shallowness” of the presentation then motivated the referees to ask some questions which were not directly related to the main contributions of the dissertation. So at least the defense really deserved the name of (self-) defense at the end.