Tuesday, November 20, 2012

About PhD Defense

Two weeks ago, I successfully defended my PhD thesis: "A Principled Methodology: A Dozen Principles of Software Effort Estimation." I have been willing to write this post right after my PhD defense, but maybe it is better that some time has passed and I had more time to think about it.

I will not  necessarily write about the details of my PhD thesis in this post (but in case you want to read the details, here is my thesis in pdf format), I rather want to discuss some of the defense presentation tips, which were particularly helpful for me. In case you are nearing your PhD defense or in case you are just starting your PhD program, I think these tips may also be helpful to you. And if they don't... Well it is not too bad either, you just lose a couple of minutes.

The first thing that is strange about the PhD defense presentation is that you are expected to present years of work amounting to a number of papers, in less than an hour of time. That is really an issue of trade-off. If you go deep into every chapter, every research idea and every algorithm, the time will most likely not be enough. On the other hand, only talking about the high level ideas and not providing details may make your research look as if it lacks technical depth. Of course there may be different blends of high-level discussions and low-level technical details that works best for different people, but the one that works best for me is to keep the discussion high-level for most of the chapters and go into complex technical details in 1 or 2 of points. That was in fact a recommendation from my advisor and it works like a charm. However, I recommend to be cautious about the choice of where you will go into these details. For example, a bad choice would be an algorithm or a derivation, which is pretty complex and fancy to discuss, but which you are not really a comfortable with. The point of the detail slides is to show that you are knowledgeable about the technical depths of your thesis. A good choice is an algorithm that you have already presented before, possibly at a conference, where you have also received questions about this technical topic.

Steering the Discussion
After the presentation, there will obviously be a Q&A session, which was in fact longer than the presentation itself for me. It is difficult to explicitly steer that discussion, since the type of questions depends on your committee members. However, some implicit details may help.

To begin with, -although you may not want to admit this to yourself- it is highly likely that none of your committee members will read the entirety of your thesis in detail. So, the material you put into the slides and the way you present it will make up a giant chunk of the evaluation of your work. It is crucially important to make sure that your slides correctly present your experiments and evaluations. One type of questions is about the technical and experimental side of your research, which are dangerous questions really. Because, in case your committee starts to think that your experiments are flawed or that there is a big need for extra experimentation, you may be looking at an extension or extra experiments.

Another type of questions will be more philosophical type of questions about what the impact of your research is, how it contributes to your domain and how future work can be built on your research. These are somewhat safer questions and it is better to steer the discussion into these questions, because after all it is your dissertation, so you are likely to be the most knowledgable person in the room about the philosophical aspects of your research. I believe the best way to get more of these questions is to provide a good motivation of your research in the presentation, a good summary at the end (possibly revisiting the motivation) and a visionary future work slide(s).

Temporal Memory
I am not really sure of the temporal memory discussion, so I will keep it brief. However, my experience from my defense presentation was that the slides towards the end of the presentation are remembered somewhat more and I have received a lot questions about the topics in the last slides of my presentation. That was also the observation of my advisor. It may not really be universal, but I wanted to mention it. A good place to test whether this is the case with your committee may be to observe it during your proposal presentation.

Blood Sugar!!
I know this sounds funny, but I it has a truth to it. People are usually crankier and are more easily agitated when they have low blood sugar, and the last thing you want is a cranky committee. It is customary to bring refreshments to your PhD presentation anyway. I recommend a lot of carbohydrates, possibly cookies (or some other desert-like thing) and/or pizza etc.

Well, the stuff that I described are my observations and experiences, which means they may not be applicable to someone else. But for what it's worth, I am also sharing my presentation below.. I hope it helps and good luck to all the fellow grad students!!