Advice to Ph.D. Students on Giving Academic Interview Talks

Raymond J. Mooney

An academic job talk is difficult because, unlike a conference presentation or normal colloquium, it requires speaking to a broader audience than just your own subarea. Therefore, you cannot assume much knowledge of the problems, motivation, terminology, and techniques specific to your subarea. However, it is also possible to present too high-level or abstract of a talk that does not convey the concrete technical ideas that make your novel scientific contribution clear, or condescends to a generally technically knowledgable and competent audience through repetition, platitudes, or over-promotion (remember, it is still a technical talk, not a marketing pitch to a company manager or venture capitalist). Therefore, it requires striking a delicate balance of accessibility and clarity while still conveying technical and scientific significance, depth, novelty, and soundness.

A rule of thumb might be that the begining third of the talk (which introduces and motivates the problem) and final third of the talk (which reviews the results and their significance and discusses directions for future research and conclusions) should be fairly easily accessible to a general computing audience; however, the middle third should contain some technical depth that should be as accessible as possible but speaks more directly to experts in the area and makes clear some of the technical details, significance, and novelty of the work.

  1. In the beginning, clearly introduce and motivate the problem you are studying in a way accessible to a broad audience. Introduce any terminology not known to everyone with a general knowledge of computing. Make it clear why the problem is important.
  2. In the middle, don't avoid giving some algorithmic or theoretical details of the work, but try to make it accessible to an audience that is generally technically knowledgeable but may not know the specific methods and terminology in your area. If you present detailed formulas or algorithms, make sure they are formally correct, use clear and concise notation, and that all technical terms are well-defined. Clearly, concisely, and precisely state any theorems or theoretical results.
  3. Perhaps pick one important aspect of the work that you can explain in some depth and detail in the time available. Don't try to cover too much and have to rush through it. It is better to clearly cover one significant aspect of the work in depth rather than try to rush through a great deal of detailed material that is not adequately explained. Other aspects of the work can be summarized in order to place this specific aspect in its appropriate context.
  4. Make sure that your specific contribution is clear from the begining and occupies a significant portion of the talk. Spending too much time reviewing and surveying what others have done and not making your own novel contribution very clear is a mistake to avoid.
  5. If you present experimental results, make sure the following are clear:
    1. What specific scientific hypothesis are you testing (.e.g. method A is better than method B according to performance metric C)
    2. What is methodology and data you used to generate the results?
    3. Make sure any graphs and tables are large, clear, and readable.
    4. Explain the specific metrics you are using to evaluate performance and why these metrics are the important ones to measure, e.g. clearly explain what the axes on graphs or rows or columns in tables mean.
    Presenting convincing experimental results on realistic data is always desirable, but make sure the meaning and significance of the results are clear and that you are comparing to the best alternative methods, not just straw men. Throwing up a bunch of tables or graphs that are not adequately explained to someone who is not an expert in the specific area should be avoided. Results should support some clearly explained scientific hypothesis.
  6. Discuss some of the limitations of the work and layout some clear directions for future research. Make it clear that you have a broad research agenda for pursuing larger significant problems in the future, not just incremental improvements to the existing work.
  7. Try to convey your excitement about the work you are doing and why you believe it is important. The audience should get the feeling that you are motivated and excited about the prospect of changing the future of computing through your current and continued research.