Reading a Research Paper

[Parts adapted from notes by Michael Mitzenmacher]

Ok, so you want to read a research paper. Where do you start? How do you make sense of this dense piece of text and math and figures.

First thing to understand is that writing a research paper is wholly different from writing any other kind of text or prose. Reading, therefore, is wholly different than the reading you’re accustomed too.

There are multiple steps

Understand why you’re reading the paper and what you need to get out of the paper. Most often you don’t need to read the whole paper. A careful reading of a paper will require several passes.

Read critically.

Be skeptical and do not assume that the writers are always correct.

Start by reading the Introduction:

  • Identify the topic of the paper.
  • Identify the contribution of the paper.
  • Identify the gap in knowledge that this contribution fills.

Then ask yourself: Are the writers solving the right problem? Are there easier solutions? Is their contribution compelling?

There may be a related works section. Sometimes the related works sections are useful and help place the work into the broader literature. Sometimes the related works are not: serve more as pointers to other works.

What assumptions are being made by the paper? Are they reasonable?

If the authors present data, did they gather the right data? There will be bias in their collection methodology. Do they account for it? Do their interpret their results correctly? Would other data be better?

Read Positively

Being critical is easy. It’s always easier to be critical than to be positive.

Try to see the problem from the writers’ point of view. What are the good ideas in the paper? Can you think of creative uses or extensions of the contribution that the authors have not yet thought of. Are there interesting generalizations and/or improvements that can be made?

Take notes

Print the thing on actual physical paper that you can hold in your hand. So much research shows that holding a physical book/newspaper/journal increases attention dramatically.

Take an actual physical pen or pencil. Hold it in your hands and take actual, physical notes on the actual physical paper. Write down comments, circle things, ask questions. This will help tremendously when you re-read the paper a few minutes from now and also when you re-read the paper a few months from now.

Identify the Way of Knowing

How do the authors conclude? What provides them the audacity the conclude they way they do?

Some papers offer new ideas, others implement ideas, other explore how ideas work in real world use cases. Still others create frameworks from lots of existing and related ideas. Some have data and statistics some use inductive proofs, other use deductive proofs.

Summarize the paper

After you’re done reading. Summarize the paper in a sentence (maybe two).

If you cannot explain something simply, then you do not understand it. [Einstein]

Usually the paper is a yes/no answer to a question. What is that question? and What is the answer?

If you find that summarizing the paper is difficult, then you don’t understand it well enough. So you’ll need to go back a read it again.

Compare it with other articles

Science is a collaborative/competitive enterprise. A research article does not exist in a vacuum; it, instead, sits with contemporary research. You will need to know the contemporary research in order to properly gauge the merit of the contribution.