Five Questions for Good Research
- What is the problem to be solved?
- Who cares about this problem and why?
- What have others done?
- What is your solution to the problem?
- How can you demonstrate that your solution
is a good one?
1. What is the problem to be solved?
Every good research project solves some particular problem. It
is very important that you be able to state your problem clearly:
in one sentence, in one paragraph, in one page. If you cannot
state your problem then you do not know what you are doing.
2. Who cares about this problem and why?
If you cannot answer this question then there are one
of two problems. 1) The problem is not important enough for
you to waste your time. 2) You have not spent enough time
considering why you are working on this.
You should be able to briefly explain
"why you should care" to the following audiences.
Many of these audiences will not understand what exactly you
are doing, but you should be able to tell them why they
- To your advisor or others in the lab
- To your peers in your research community
- To other computer scientists
- To other colleagues in the university
- To your Mom and Dad
You should also clearly understand the set of people who
have the problem you are trying to solve and be aware of
what their needs actually are. If you do not know the people
you do not truly know the problem.
3. What have others done?
Good research never exists in a vacuum. It builds upon or
responds to the work of others. You must be able to set your
work in the context of other work. There are two aspects
of prior work that you will need.
- Who has previously attempted to solve this problem?
Why did they not succeed or how is your problem a
different, special case or more general case of the
problem that they solved?
- What concepts, techniques, mathematics or software
have others developed that can form a foundation for
4. What is your solution to the problem?
This question is answered in two ways depending on your stage
You should try to sketch 3-5 key insights that you believe
will allow you to solve your problem better than what has
been done before.
You should have 3-5 key ideas that address the problem. You
should be able to state these ideas in 1 sentence, 1 paragraph
or 1 page. You should be able to describe these ideas in
sufficient depth that others who are knowledgable in your
area could reproduce the work.
5. How can you demonstrate that your solution is a good one?
Just having done an interesting thing. You must be able to
show that your thing is better than what has gone before.
- Your problem may have no prior existing solutions.
therefore you must show that your solution really solves
- Your problem may have existing solutions in which
case you must identify clear criteria for comparing two
solutions. You must show a strategy for evaluating
criteria. It is easiest if there objective
comparison metrics. You must show that your criteria and
your evaluation are valid. Does your criteria really
reflect the needs of your problem and your users?
Does your evaluation strategy fairly evaluate the
criteria or are there statistical or methedological flaws?