A thesis proposal is a contract

[article index] [] [@mattmight] [+mattmight] [rss]

Of the three official milestones1 in Ph.D. school--qualifying exams, thesis proposal and thesis defense--the trickiest is the thesis proposal.

Qualifying exams are just exams. Students can beat them using the same tactics they used for exams as undergrads.

A proper thesis defense should be a rigorous formality. No advisor should ever let her student stand for a defense unless the advisor is convinced the student will pass.

Thesis proposals, on the other hand, rarely pass without an objection requiring a modification to the proposal.

Students tend to invert the importance of the proposal and the defense: they see the proposal as the formality and the defense as the challenge.

Faculty, however, treat proposals like contracts. When faculty sign off on a proposal, we are giving up some of our rights to object later on. We are agreeing that if a student does X, Y and Z, as outlined in the proposal, then we shall grant a Ph.D.

We are never going to yield such rights passively.

All good thesis proposals contain two ingredients: a clearly-defined thesis, and a specific plan for demonstrating that thesis.

Everything else in the proposal (related work, prior work, challenges) exists to support the plausibility of the thesis and the plan.

Done properly, a good proposal grants immunity against remarks like

  • I thought you said you were going to do X,
  • I'd like you to also do Y before I sign off, and
  • So what?
during a defense.

A thesis proposal is an opportunity for students to obtain protection and assurance. It should restore some of the certainty lost in the middle years of a Ph.D. Too few students take full advantage of this opportunity.

If you need advice on how to do a thesis proposal, read on for guidance.

Thesis statement

A thesis is a single sentence.

More precisely, it is an active, declarative, defensible sentence.

The thesis is what the dissertation will exert its mass to defend.

The thesis proposal document and the proposal presentation should place the thesis statement toward the front, so that while reading and listening, the committee can determine whether the thesis is defensible, and whether the proposed research, if successful, would constitute a valid defense of the thesis.

When a student doesn't provide an explicit thesis statement, each committee member will infer a thesis statement that aligns with his or her prejudices, and then, some day, the student will have to defend six theses instead of one.

My advice is to make the statement as short as possible: how much fat can you trim from it before it's too general too defend with your work?

The thesis statement answers the question, "What did humanity learn as a consequence of this dissertation?"

It's useful to carefully define each term in a thesis statement after it's given, just to make sure the student and the committee are in total agreement about what the thesis means.

Example: My thesis

My thesis statement was, "Environment analysis of higher-order languages is novel, feasible and useful."

This thesis segmented my dissertation into four parts: related work, theory, experimentation and application.

  1. Related work defends novelty.
  2. Theory and experimentation defend feasibility.
  3. Application defends usefulness.

My thesis statement doesn't say anything about what technical mechanisms I used to prove environment analysis feasible. (I developed abstract frame strings, abstract counting and abstract garbage collection to support my thesis. But, those were just the means. Environment analysis was the end.)

The weakest part of my thesis statement is the claim of usefulness.

Of course, even if I took out that claim, my committee would still ask the questions, "So what?" Putting it in my thesis forced me to explicitly defend against that question.

Utility is subjective, which meant I had to find an application of environment analysis for each of my committee members.

Ultimately, I detailed 16 distinct applications of environment analysis, so that if a committee member wanted to argue it wasn't useful, they had to argue that all sixteen applications were useless.

None of my committee members were willing to take up the challenge.

The modifier novel is also unnecessary. Committees will look for novelty whether it is claimed implicitly or explicitly. Again, putting it in my thesis forced me to defend novelty in my dissertation.

Plan

Students need to realize that a proposal is a contract.

The plan details the conditions of that contract.

If a student words the plan right and gets it approved, her defense will go smoothly.

If she leaves the plan vague or inspecific, she leaves herself vulnerable to the committee's interpretation of her plan.

Nothing of importance should ever be left to the mercy of a committee.

A good plan contains a fictional schedule--a list of remaining milestones and anticipated dates of completion. The dates don't matter all that much. They just have to be there. Relative spacing between dates signals the estimated size of the challenge for each milestone.

If the plan contains a claim to be validated, it needs to explain how the student will conduct validation of that claim.

A good plan also contains contingencies. What if a claim ends up being false? What if it turns out to be infeasible in time or in cost to perform the evaluation of one of the claims? A good plan is not a sequence, but a tree. The leaves of the tree form a spectrum from "best possible outcome, give me a Ph.D. and a professorship" at one end to "back to the drawing board" at the other.

A plan should contain the possibility of failure.

Real research is inherently unpredictable, and failure is always a possibility. If failure is not possible, it must not be research.

The proposal needs to create the impression that failure is unlikely.

A good plan also provides the criteria for recognizing the completion of a milestone, e.g., submitted for publication, accepted for publication, survey completed, chapter written.

Supporting ingredients

The remainder of a proposal exists to support the thesis and the plan. The other three components in a good proposal are (1) a survey of related work, (2) a summary of prior work by the student and (3) a review of research challenges.

A survey of related work supports the novelty of the thesis.

A summary of prior work gives the committee a sense of what the student is capable of, which supports the plausibility of both the thesis and the plan.

A review of the research challenges and proposed circumvention strategies supports the intellectual merit of the thesis.

Good proposals give the impression that between one-third and two-thirds of the work remains to be completed.

Thesis proposals claiming that all of the work is already completed will be interpreted (rightly or wrongly) as arrogant, and trigger intense scrutiny.

It's important to propose before all of the work is finished.

If it's truly all done, a student should pretend the last third of it isn't.

Wielding the proposal

At a student's defense, she should put the thesis statement up at the start, and then circle the path through the plan she ended up taking.

Psychologically, it's going to be hard for the committee to say she must do more work after they're reminded of the plan they agreed to.

Pragmatics

A good thesis proposal document can be structured like a proposal for NSF funding: a one-page proposal summary, a fifteen-page proposal description and an extensive bibliography.

Proposals longer than fifteen pages are unlikely to be read in full.

The presentation itself should be about 30 to 45 minutes. As proposals drag on, busy committee members get crankier. Cranky faculty members revert to their basest instinct: being argumentative. Running long is a well-tested strategy for having extra conditions placed on a proposal.

As a rule of thumb, each 5 minutes past an hour adds 3 weeks worth of work to a Ph.D.

Related posts


[1] The three unofficial milestones are the first publication, the second publication and the third publication.


[article index] [] [@mattmight] [+mattmight] [rss]