One Strategy for Writing a Research Statement

If you're in the market for an academic job, chances are that you'll have to submit a research statement of some sort. These critters offer you the chance to distill your life's work and dreams into a short essay that can only hurt your chances of employment. Since you probably just finished your thesis where length is a virtue writing this statement can be daunting. While examples of research papers abound, there are relatively few research statements on the web. None of those were peer reviewed. Most of them were written by fresh Ph.D.s who found themselves in exactly your position.

To help you out, I'm going to describe a fairly specific structure to follow. This may or may not fit your situation, and it may not generalize beyond computer-related research. I developed the structure out of need by analyzing statements from people who (a) have jobs and (b) work in my general field. Recently IEEE Intelligent Systems ran the feature "AI's 10 to Watch" where each person provided a 5 to 10 paragraph research statement suited for a relatively general audience. These vary in quality and message, but more importantly they have commonalities that one can distill.

The outline below fits my own work fairly well, and it may cover your case too. If not, then you may find a better model in one of the other narratives based on the degree of impact and novelty of your own research. Additionally, you can read the previous installment from 2006.

The Outline

  1. Define your general problem area. If it's a well-known area, like natural language processing or machine learning, then identify what you think is the primary question and its merit—that is, take a stance. If it's a new research domain, motivate it. If it's really new, see the first 3 paragraphs of Levin's statement from the IEEE article.

  2. Narrow your focus to a particular aspect of the domain (e.g., scope-limited natural language processing, path planning in unknown places, role of social networks for semantic web), state your own motivation for doing so, and note the limitations. You can also describe Big Problem + Challenge + Approach at a high level. The details of this paragraph depend on how familiar your problem is. Think in terms of a subtask, but not the solution. For example, this paragraph might talk about "transfer learning as a problem in machine learning" but it would not talk about "analogy for transfer learning", which comes next.

  3. Announce your secret weapon. What do you have that others don't? Why can you address the challenge above when others have failed or even overlooked it? This could/should be some technology that you built. Don't be too technical here, because nobody cares if you used Markov-Chain Monte Carlo, Gibbs sampling, etc. If your research has already had an impact, mention it at the end of this paragraph (e.g., "Hundreds/Thousands/Millions of people/researchers use this all the time!") Otherwise, mention how this tool helped you address the challenge from paragraph 2. The tone of this paragraph has to be just right. See the statements by Morency and Cimiano and compare those to Sudderth and Milch. Luis von Ahn has two "secret weapon" paragraphs, but his research statement differs from the others due to the practical focus and large impact of his work.

  4. State your current focus. What's the big unaddressed problem? Why hasn't it been addressed? (Hopefully because your technology wasn't around.) How will you address this problem? This should be particularly passionate, and it can help to include a bigger or more general long term goal.

  5. End on a strong note. This paragraph (or the last one for longer statements) was the most variable in the examples. Sudderth's statement paints a rich vision of the future. Other statements cover the "big question" in paragraph 4. In that case, the last paragraph reflects "my approach" and ends with a sentence that claims an expansive general impact or that broadens the focus to the level of the first paragraph. Do not end with "My domain is great/relevant!" Consider Mika's last paragraph as a positive example.

Overall, I think that 5 paragraphs is sufficient for a starting point. You can always add more "Big Problem + Challenge + Approach" paragraphs if you've done two or more great things, but brevity is of great value. If the statement's only one page, people may actually read it.

Postscript: I'd love to hear your opinions on this strategy or if you have one of your own. If you're wondering, my own statement (updated 11/20/08) diverges from the outline to a slight degree. Feel free to read it and decide for yourself if this approach works.