Wednesday, June 22, 2011

Applying for Jobs: the Job Talk

This is the fifth post in a series on applying for faculty positions. It is written by Jeff Phillips, frequent guest blogger, who will be starting as an Assistant Professor at the University of Utah starting Fall 2011.



The final component of applying for jobs is the job talk. It will usually be about an hour, but may vary so ask. Usually this means you can talk for about 58 minutes, before they want to stop you for questions. But again, its best to confirm this, try to ask your host specifically about this. Also ask about how frequently you should be expected to be interrupted with questions during the talk. Of course, this may vary greatly from talk to even at the same place (as I have seen in my experience). But, from my perspective, many instances where the speaker was delayed for more than 5 minutes due to questions in the talk was the fault of the speaker. This usually occurred by leaving ambiguities in the talk, by failing to answer questions clearly and concisely (this may lead to more questions), and by not taking charge and insisting on taking longer questions off-line.
The most important failure (ambiguities) will hopefully be avoided by taking the steps below.



First, what should you talk about?
  • Talk about your best work!
  • Talk about your work that appeals to a broadest audience and that you can convey the main ideas of most clearly.
  • Talk about the project you are most well-known for, this is probably because its your best received, and likely your actual best work even if you have some "personal'' connection to some other topic.
  • Talk about the work that is most likely to get you hired. This last one can be tricky, but try to ask your host or anyone else you know at the institution what specific type of person they are looking to hire. No one will fit a mold perfectly, but make sure the faculty there don't have to squint too hard to see you in fit in that mold.
If it is the same topic that fits all of these criteria, then you have a much easier task. Otherwise, you may need to compromise on some these. Try not to sacrifice strength of work and clarity of key ideas. If you are not sure which topic is your strongest, ask around.



What should go into talk? Not too much.
Present at most 2 or 3 main ideas. Most papers have one main idea, even if there are several smaller ideas in the paper. The lasting contribution is usually one key insight, (even if it took a lot of hard work and many other technical details to get everything to work). Try to abstract away those technical details, and convey the main new ideas your work added. You can mention that there were difficult technical details, but do not describe them!

So 2-3 main ideas equates to 2-3 papers. Perhaps you have a series of papers that layer on interesting ideas towards a single goal. In this case, you can possibly convey more, but do not try to cram more at the expense of not making clear what the main conceptual contribution is.

It is also helpful to have the talk tell a story, to have a single narrative. But, if you have two pieces of work that are both your strongest works and easiest to express the key contributions, then give the talk in two parts and choose to convey better work than to give a cleaner, but weaker story.

If you are a theoretician, give at most 1 proof. (Unless, with possible exception, you are being hired into a strong theory group and they alone make the hiring decision, then, maybe, give 2 proofs). Most people will get lost in the details. I gave one proof sketch, but "intuitively explained" why several things were true, usually each in one to two sentences. I am guessing for non-theory people, there is an equivalence between number of "proofs" and number of some other technical descriptions. The point is, it might be good to spend 3 minutes at some point showing off your technical prowess, just to try to convince the audience you are an expert in an area - but there can be better ways of doing this than diving so far into the technical details that you somewhat intentionally loose most of the audience.

You should aim to teach the audience something. That is, go in a bit more depth for some general technique in your field that you feel is not well-understood, and would serve the general audience to know. This does not need to be explicitly your work (and often will not be), but you may have extended this work. If you did extend it, the audience will appreciate it much more if they understand the importance of the general version. In this "teaching segment" the goal should be to allow the audience to understand how to operate some heavy machinery (at least in principle). Spend time developing simple examples that clearly demonstrate how it works and its power.
This segment is also important in demonstrating your teaching abilities. If someone leaves your talk comes away feeling they learned something, even if it was not your specific contribution, then they will have a positive impression. I know when I spend an hour in a talk, and don't come away with this impression, I am very disappointed in the speaker.



How should you structure your talk?
Spend the first 10 minutes motivating the entire area you are working in, and then the set of key problems in the area. Start at a very high level, and use well-planned examples to zero in on what are the major hard problems in the area, and why they are the critical remaining component in advancing the sub-area. This should allow you to outline the rest of the talk by just saying which of these problems you solved (and then spending 40 minutes explaining how).
This first 10 minutes of motivation can be modified if you give job talks at several place without changing much of the forth-coming meat of the talk. This may be necessary to paint yourself in a few slightly different roles depending on what each university/lab is looking to hire. Minor changes in the set up, can give the talk a very different flavor, perhaps focusing on different aspects that potential collaborators at each institutions could appreciate as fodder for possible proposals they could write with you.

Then spend about 15-20 minutes each on covering your contribution in 2-3 core problems. Probably don't spend more than this on any one topic, since if you lose some audience members, you can re-captivate them for the next section. And any less than this, you probably will not be emphasizing the contribution enough, and it you don't feel it deserves that much, then your talk might be cleaner if you did not waste the time to mention it.

Within each of these sections, explain the core problem in more detail, explain what other approaches are available and why they do not work. This should lead naturally into what you did. Do not just show results saying that your approach was better, make sure to explain the key idea of why your approach was better. What made you approach succeed when no one had before? If an audience member cannot immediately answer that question, then they may likely come away with the impression that you did not do anything interesting, other than tightening the existing bolts. You can conclude each section by mentioning extensions to this work, either your or others. These extensions are especially useful if they demonstrate how your key idea was integral in advancing the field.

Finally, there is likely other work that is not one of they 2-3 key contributions that you have done. You probably want to talk about it more than you actually should :). I dealt with this in two ways. Given than I had described a set of an important problems in the first 10 minutes with several sub problems, after a couple sub-problems I described my contribution for, I listed on a single page all of the other work I had done in this related sub-area. This allowed me to mention it and also have it in context of the field I was overviewing.
The other way was just to have 1 or 2 slides are the very end of the talk listing other work I had done. Its good to show you are broad, and that you have completed a lot of work. But its also easy to see this from glancing at your CV. So in these 1-2 slides attempt to cast each other project you did not talk about in relative comparison to the work you did talk about. If you mentioned 2 projects, and you had 4 other ones of similar proportion, try to convey this. This is more informative than CV paper-counting.

Conclude your talk with your vision for the next 5-10 years worth of research. I felt it best not to focus on specific problems, but on how the work I had done would have an impact down the road. Anyone can do work in a certain area or solve specific problems, but if you convince the audience that your work will be important for a lot of important future work, then not only did you have the foresight to produce that work, but you are well-positioned to continue to contribute in the field.



Finally, practice, practice, practice!!! I think I practiced my talk at least 50 times. Thats 50 hours of just speaking. When I mentioned 58 minutes to speak earlier, I meant more-or-less exactly 58 minutes (plus or minus 15 seconds). If you practice enough, you will know exactly how long the talk takes as a whole, and how long each sub-section takes. If you get a lot of questions in some early part and lose 3 minutes, you will know where you can make it up in a later part.

Also, practicing it will help you realize what is explained well and what is not. If you repeatedly find yourself explaining a slide and wishing you had a figure to help explain that, then add the figure. If there is another slide that has a figure that you don't really need, then remove it.

I only had the opportunity to practice in front of an audience (other than at actual interviews) twice. So, I spent a lot of time sitting in a room by myself talking to myself :).

So, the key take-aways are: (1) motivate the field to convince the audience your problems are important, (2) make sure you convey 2 or 3 main conceptual contributions, focusing on key new ideas, (3) teach the audience something, (4) demonstrate vision for the future, and (5) if you practice enough and keep these in mind, you will definitely give an excellent talk.

4 comments:

  1. Thanks for this post. I have a minor comment -- I think the amount of practice talks needed can vary drastically, depending on the person. I could not imagine giving 50 practice talks. It's probably better to practice until you are comfortable with the talk -- you can "defend" the choices you made (why do it this way, and not that way?). Also, a talk can get worse over time, unless you change around things to keep it interesting for yourself.

    ReplyDelete
  2. I think that 10 minutes motivating the area is not sufficient. 15 minutes is better. For many people in the audience, this is all they will get from your talk, and it is the most important part. It is well worth going slowly here. Step back to find the "obvious" motivations that you might not have ever said aloud before.

    I also agree with the previous comment. Overly polished talks can easily become boring, to the audience and potentially to the speaker.

    ReplyDelete
  3. Thanks for the comments. For my first interview, I had not yet given that many practice talks and it went fine, but felt the talk definitely improved as I practiced more. I also changed the intro (and as a result, the timing of everything else) for each time I gave it.

    The point of doing practice talks is not to make it robotic, but to put yourself in different situations. For instance, practice doing the last section in 10 instead of 15 minutes. I have seen talks that did look overly rehearsed, but if you keep this in mind as you are practicing, you will eventually surpass this stage and it will again seem natural and not rehearsed.

    And again, to each your own, but I figured this talk was pretty darn important, and 50 hours was a small fraction of all of the time I had invested in grad school and as a postdoc leading up to this moment.

    ReplyDelete
  4. The new background in this blog is terrible - it slows down the mere reading of it. Please change/fix it. Thanks!

    ReplyDelete

Disqus for The Geomblog