What's in a Thesis Proposal Spencer Rugaber March 4, 2009 I. Introduction Because of the greater uncertainty involved, students often agonize more over doing a thesis proposal than they do over actually completing their dissertation. In an effort to reduce some of the uncertainty, this memo contains my suggestions for the content and format of a thesis proposal in Computing. It is a companion to my piece on the dissertation itself: THOUGHTS ON THE STRUCTURE OF CS DISSERTATIONS, which can be found at URL http://www.cc.gatech.edu/fac/Spencer.Rugaber/thesis.html. II. Content The purpose of the thesis proposal is to convince the reader that the student is sufficiently prepared to enter candidacy and produce a dissertation. To do this, it must give evidence that the student understands the field, has recognized and defined a worthwhile problem, and has an approach and plan for solution. As a demonstration of this, I look for the following items in the proposal: + Problem motivation + Related work + Problem description + Research questions + Thesis statement + Solution approach + Evaluation criteria + Plan of Action A. Problem Motivation The problem that you are trying to solve needs to satisfy several criteria: it must be relevant, it must be significant, it must not have been already solved, it must be well-defined, its solution must be feasible, and it must be neither too large or too small. 1. Relevance, Significance, and Timeliness You are working in a field, Computing, and within an area of Computing such as compilation, or systems. Areas are not arbitrary--they are defined by a class of related problems such as language translation or resource allocation. As such, your topic normally will belong within some field. The field defines a vocabulary and, more importantly, it lays out the important or defining issues for itself. Hence, it is important that you describe your problem using the vocabulary of the field. Moreover, it should address an issue that the field deems important. Interdisciplinary theses are possible and, in fact, the College of Computing encourages them. However, they increase the burden on the student to justify relevance and significance to multiple fields simultaneously. One approach to insure relevance and significance is to pick a problem that lots of people are working on. There is a danger here, however, that someone else solves the problem before you do. This is called being "scooped". A remedy is to "leapfrog" the competition; that is, to assume that someone will solve one of the field's outstanding problems soon, and that their solution will raise a further problem that you, with your great foresight, will have a head start on solving. The risks of this strategy are obvious. Another, less common, difficulty is to pick and solve a problem before the world is ready for the answer. PicturePhone, a way to send video over telephone lines, was solved in the early 70's, but the enabling conditions for its adoption still aren't all the way here. 2. Specificity One thing that your committee will look deeply at is how well you have defined your problem. One way you must deal with this is by having explicit evaluation criteria and exit conditions. These, in turn, require you to put a lot of effort into rigorously, possibly formally, defining the problem. Student often find that when they try to do this, the problem either becomes too hard or it disappears entirely. 3. Feasibility and Size As a general rule, graduate students are ambitious and often describe a research plan that is too large or too difficult. In fact, the most commonly occurring comment that a committee gives to a student upon hearing a proposal is to pare it down and focus on the essentials. I see nothing wrong with this, and, in fact, have told students to describe an overly ambitious research agenda in order to give the committee something to recommend :-). B. Related Work When you have reached the stage where you are writing a thesis proposal, you should have a firm understanding of the work of other researchers in your field, both past and current. In fact, you should be one of the world's experts in your area. If this seems daunting, consider it as part of the price that you have to pay to earn that description. To obtain that understanding, you should have both a deep and broad familiarity with the research literature. By "deep", I mean that you should have read all of the papers in your area to the extent that you can understand their contribution, its weaknesses and its relevance to your work. By "broad" I mean that you should also be familiar with work in related fields, to the extent that you have read the key papers and survey articles. There is a danger that the Related Work section of a proposal (or a dissertation, for that matter) can give the impression of being boilerplate. That is, that it was independently created and appended to the proposal because the proposal has to have a Related Work section. Actually, discussing related work is a part of motivating the research. Recall that part of problem motivation is relationship to a field. And this relationship is justified by the historical record of important results in the field. Ideally, the Related Work section should be distributed into the Motivation and Problem Description sections as exposition and citation of the fields defining results. However, exact placement is less important than demonstrating that the student understands what is truly relevant and significant. C. Problem Description Most of the work preceding a thesis proposal goes into defining a problem and demonstrating a feasible approach to its solution. Consequently, an important part of the proposal is a precise presentation of what that problem is. There are two components of this presentation: a set of salient research questions and a thesis statement. D. Research questions A Ph.D. dissertation is an argument or demonstration of a thesis. The argument has a structure, called its rhetoric, that consists of a series of assertions, each individually justified, that when taken together imply the thesis. If we use the analogy of a mathematical proof, the thesis statement takes the place of the theorem being proved, and the answers to a set of research question takes the place of the assertions. E. Thesis statement The work preceding a proposal involves detecting and defining a problem that is worth solving and making sufficient progress in solving it that the student can convince the committee that the solution can be completed in a reasonable time. An excruciating impediment to success with the first part of the above is to construct a thesis statement. The exercise is difficult because it requires the student to say in a single sentence what it is that they will be trying to convince the dissertation reader of. The single sentence requirement forces the student out of the morass of details with which they have been concerned to a high level of abstraction. And, recalling that a "thesis" is an argument, the thesis statement requires the student to express their proposed results in positive terms with appropriate qualifications. A generic example of a thesis statement goes something like the following: The FOOBAR algorithm will be able to solve the MUMBLE problem in time proportional to the log of the size of its input, thereby improving the current best solution. F. Solution Approach You must be able to convince your committee that you are on the right track. Part of this process is to define the problem and part of it is to motivate its importance. But the most important part is to demonstrate that your approach has a real possibility of succeeding. The most direct way to do this is to point to already published results. Specifically, your proposal should include a fleshed-out summary of your preliminary results and citations of your publications. Within the College of Computing, there is a tendency for proposals to come relatively late in the research process, sometimes occurring only a few months prior to the defense. In my mind, this defeats the purpose of the proposal, in which the committee can provide significant input to guide the direction of the research. Hence, you need to find the right moment--after you have enough results to demonstrate feasibility but not too late to benefit from your committee's advice. As stated above, a dissertation is an argument. Arguments are structured as a sequence of supported points. So one important part of the proposal is to provide a sketch of the argument. One way to do this is to provide an annotated outline of the dissertation itself. This has the added benefit of getting the student started on the process of writing the dissertation itself. G. Evaluation Criteria Ultimately, the goal of an advisor is to train a PhD student in terms of his or her research skills. Essential skills include critical reading, problem recognition, writing, and, most importantly, research evaluation techniques. These techniques will vary with the research field. For example, theory students will have to devise formalism and express arguments mathematically. Human factors students have to devise and conduct experiments and perform statistical analyzes. Engineering students usually provide validation for their work in terms of a prototype and one or more case studies. All of these skills must be learned and practiced, and no proposal is satisfactory without a well-defined evaluation plan. The primary quality criterion for Computer Science research is not some fuzzy excitement generated by an idea, but rather the extent to which the idea can be generalized and applied. And the best way to demonstrate this quality, is to have a convincing validation. H. Plan of Action A thesis proposal is just that--a proposal. Part of what is being proposed is an idea of how to solve a problem. But the committee also needs to determine the feasibility of completing the proposed work in a reasonable amount of time. Like any planning effort, the task to be accomplished should be broken down into activities. Dependencies among activities should be recognized, and time estimates should be prepared. A sketch of this plan should be included in the proposal III. Summary The proposal process should be one of convincing your committee that you have defined and circumscribed an important problem, that you have a promising new approach to solving it, that you know how to determine whether it works or not, and that you have the requisite skills to do the work. The proposal itself should present this evidence in a coherent fashion in a way that simultaneously demonstrates maturity and solicits suggestions for improvement.