Being a Discussant: Reading Someone's Paper Properly

You have been asked to discuss a paper. In most fields nobody explains what that means, and the result is improvised — either a summary of the paper the audience just heard, or a list of objections delivered to a person standing three feet away.

The role has a purpose, and it is one of the few genuinely useful things at an academic conference. For a concrete operational example of how a workplace tool approaches a related coordination problem, see this guide.

The role varies substantially by field. In some, discussants are standard and prepared; in others the term barely exists. What follows describes the version common in economics and adjacent fields. Adapt to what your conference expects.

What a discussion is for

Three audiences, three purposes.

For the author: the most careful reading their paper will get before review. A discussant who has read properly gives them something a referee report cannot — a reaction from someone thinking aloud, in time to act on it.

For the audience: context. Where this sits in the literature, what is genuinely new, what to be alert to. Many people in the room have not read the paper and are deciding whether to.

For the field: a public model of how to engage with work seriously. This matters more than it sounds, particularly for the students in the room.

What it is not for: demonstrating that you are clever, or that you would have done it differently.

Preparing

Read the paper properly, well in advance. This is the whole job. A discussion prepared on the flight is visible to everyone including the author.

Read it twice. Once to understand, once to interrogate. The second reading is where the useful comments come from.

Work out what the paper is actually claiming, in one sentence, in your own words. If you cannot, that is itself the most important comment you have.

Identify the two or three things that matter. Not everything you noticed — the things that would change how the work should be read or what the author should do next.

Ask the author what would help, if you can. Some are looking for help with identification, some with framing, some with where to send it. A short email a fortnight before makes the discussion far more useful.

The structure that works

Ten minutes, roughly.

One minute: what the paper does. In your words, generously. This demonstrates you read it and orients the audience.

One minute: why it matters. The genuine contribution, stated clearly. Do this properly even if you have substantial criticisms — a discussion that opens with objections tells the audience nothing about whether the work is worth their attention.

Five to six minutes: two or three substantive points. Not a list. Each one developed enough that the author can act on it.

One minute: what you would like to see next. Framed as an opportunity rather than a demand.

Leave time. Discussants overrun more than presenters, and it eats the audience discussion, which is the part the author benefits from most.

What makes a comment useful

Specific. "The identification is unclear" is not actionable. "The variation you rely on could also reflect X, and here is why that matters for your interpretation" is.

About the paper's own goals. Criticising a paper for not being a different paper is the most common failure and the least useful. Take the question as given and engage with whether it is answered.

Answerable. A comment the author can do something about beats one that would require a different dataset that does not exist.

Honest about its weight. Distinguish "this would change the conclusion" from "this would strengthen the paper" from "this is a presentational suggestion." Delivering all three in the same tone means the author cannot prioritise.

And say what you found convincing, specifically. This is not politeness — it tells the author which parts to protect when revising.

The two failure modes

The summary

The discussant restates the paper. The audience has just heard it. The author learns nothing. Ten minutes gone.

Usually a symptom of insufficient preparation — summarising is what you can do having read once.

The demolition

Every weakness, delivered publicly, to someone who cannot respond at length.

This is a status performance rather than a discussion, and it is recognised as such by everyone in the room including the people who are impressed by it.

The test: would you make this comment, in this way, in a referee report where nobody knew it was you? If the public delivery is the point, it is not a discussion.

Harsh criticism is sometimes correct. Deliver it as you would want it delivered — specifically, with the reasoning, and without the audience as a party to it.

When the paper is weak

It happens, and improvising is worse than preparing for it.

Be honest without being cruel. The audience can tell when a discussant is padding, and pretending a paper is strong helps nobody.

Find what is genuinely there. A weak paper usually has one good idea, and identifying it is a real service.

Focus on what would make it work. "The current design cannot answer this question; here is a design that could" is criticism and it is also help.

Keep it shorter. A long discussion of a weak paper is uncomfortable for everyone.

Say the serious thing once, clearly, and do not return to it.

Practical notes

Send your comments to the author afterwards, in writing. Ten minutes cannot carry everything, and the written version is what they will actually use.

Do not use slides, in most fields, unless you need to show something specific. A discussant with fifteen slides is giving a second talk.

Do not present your own related work. It happens constantly and it is not what you were asked to do.

If you cannot read the paper properly, decline. A discussant who has not read it is worse than no discussant, and declining early lets the organiser find someone else. See declining academic requests.

Why it is worth doing well

It is the best training available in reading critically, and it transfers directly to reviewing. See how to review a paper.

It is visible. People notice a good discussant, and it is one of the few ways to build a reputation for judgement rather than for output.

And it is reciprocal. The quality of discussion you receive is a function of the norms in your field, and those norms are set by what people do.

The short version

Read the paper twice, well in advance.

Two or three substantive points, not a list. Specific, answerable, and about the paper's own goals.

Say what is genuinely good first, and mean it.

Send the full comments in writing afterwards — that is what the author will use.

For an independent authoritative reference, consult the MIT Communication Lab for academic communication guidance.