(Adapted from Ram's essay of the same title, which he claims was paraphrased from memories of an earlier essay (now lost) by McDermott. This version is from autumn of 2011, used in a course taught by McDermott.)
Many students present a paper, especially one authored by someone else (the situation I'm going to assume in most of what follows), by talking through it section by section or page by page. The presenter reads out the definitions and points the audience to the figures. They work through the paper linearly, taking great care not to miss anything that the author might have written that might possibly be relevant. There are two problems with this approach:
The audience doesn't want to hear the paper read; they are perfectly capable of reading the paper themselves, and, we hope, have already done so. What they would like is some illumination cast on it, a point of view that calls attention to what's important in the paper, good or bad.
Here is the approach we recommend.
Read the paper carefully. Most papers in computer science make and back up some technical claims. It's unlikely, although possible, that some fallacy or overlooked case has made the results false, irrelevant, or useless. It's somewhat more likely that the results just aren't novel. In some subfields, such as artificial intelligence and systems, there is the further issue of empirical evidence for claims about how well an algorithm performs under realistic conditions. It has been known to happen that the evidence is flimsier than the author realizes or admits. In AI, there are even papers that wander into philosophy, and there may be questions about whether they should have wandered and if so whether they have fallen into quicksand.
So some papers have so much wrong with them, or so believes the presenter, that there is almost nothing salvageable. Most, however, have at least one redeeming idea, some nugget of information that is the reason why they deserve to be read and assimilated.
The presenter's primary decision is whether the paper has some good ideas or whether it belongs in the recycling bin. Let us repeat that, in our opinion, very few papers fall into the second category.
The next step is to decide which idea or cluster of ideas is the central raison d'etre of the paper: the most novel, the most clever, the most unfairly rejected by the community, etc.; or which fallacy or circumstance is the main reason why the paper is a clunker. We'll use the word "central issue" to describe the hero or villain of the piece in these senses. Write down this idea, preferably in your own words, and a one-line justification for why this idea is the best one. (This is for your use, probably not to be shown to the audience.)
Now comes the crucial step: Figure out how to get your audience as quickly as possible to the point where they can understand the central issue of the paper. Withhold editorial commentary for now. Take the author's point of view.
Next, if necessary, elaborate the idea and fill in the details. Explain things like how the idea came about, how it was fleshed out in the paper, how it was proven, what benefit it had or might have, what difference it made or might make, what alternative ideas might have been pursued instead, and so on.
But don't overexplain! There may be some key lemmas that are needed to understand the central issue, but don't go through their proofs unless (a) the central issue is the novelty of a proof; or (b) the central issue is that one of the proofs is faulty. You've got to leave enough time for the next step
State your evaluation of the central issue in the paper and justify it. If there is a nugget of wisdom, explain why it's wise (at least a little bit). If there is a big blunder, explain why it's wrong. Feel free to exaggerate a little here. You want to provoke some discussion, so mention why this paper's idea is better than idea X, the one that community Y takes as central dogma. Or explain why the mistakes in this paper are typical of people in community Z. Hopefully at least one of the people in the room belong to community Y or Z.
Leave some time for discussion. Hopefully you have stimulated some.
What changes if the paper is your own? If it's something you wrote years ago, then not much. In fact, you can think of your younger self as a different person from you, and proceed as if someone else wrote the paper. But if the paper is recent, then, unless you're a masochist, presenting the paper as trash is not a good idea; seek therapy instead. Other than that, most of the recommendations above are still worthwhile, except the bit about exaggerating the virtues of the paper. Chances are you'll do this unconsciously, so don't compound the offense by doing it consciously.
There is another group of issues that arise if one of the authors of the paper is a former student of the institution at which you are now a student or faculty member, or is your current or former advisor, or is a former student of yours, or is your advisor's worst enemy, or has some other fraught relationship with you. The only advice we have is, if at all possible, to pick a different paper to present. To say more would get into political issues that are beyond the scope of this essay.