Group discussions, critiques, and in-class discussions
Group discussions
In groups of about 3-4 people, meet for 1 hour after reading
the paper but before writing your critique. Spend some time
(perhaps 20 minutes, depending on the paper)
making sure everyone understood the important points, and spend
some time starting to discuss the points related to the critique
(contributions, strengths, weaknesses).
A paper's discussion leader should also use this time to
get feedback from his/her group, especially on his/her intended
discussion questions.
Critiques
After the group meeting, write a critique that has five short
paragraphs:
- Problem that the paper is solving (and why is it important and/or hard and/or why doesn't prior work address
it?).
Try to identify the key technical problem being solved.
- Key insights / contribution / strengths (in your own opinion; don't regurgitate what the papers says).
What are the key technical insights and contributions of the paper?
- Weaknesses / fallacies / flaws (not the obvious ones or the ones mentioned by the paper, but rather the ones
you think are the most serious / most consequential, and why).
- Unsolved problem(s) / opportunities for future work.
- Your difficulty/ies in understanding.
Although each paragraph can be short, avoid paraphrasing the paper
and avoid superficial statements. For example: "The authors propose
an approach to detect atomicity bugs, and they evaluate its performance
and accuracy." This sentence doesn't really say anything unique to one paper.
What's the key aspect of the approach and the key insight that makes it work?
What is the (single most important)
outcome of the performance and accuracy results, and why?
Some additional advice:
- Try to say non-obvious, interesting things.
That may seem like generic advice, but it's a useful filter to apply
to any given sentence, i.e., "Is this sentence going to be interesting
or non-obvious to someone who has already read the paper (but perhaps
not thought about it in depth)?"
- Try to convince the reader that you've not only read the paper, but that you
understood (at least some parts of it) at a fairly deep level, and you
really thought about the papers ideas and contributions.
- You can assume your audience (me) has read the paper. This assumption
should help you skip unimportant sentences.
- Don't try to cover everything. Depth is more important than breadth.
- Try to figure out what are the paper's real contributions,
strengths, and weaknesses—which might be different than what
the paper says about itself. Take risks; it's okay to be "wrong."
- Take risks. It's better to be wrong than boring/unimaginative. :)
The critique for a paper is due on the day before we discuss the
paper in class. As part of your critique, include your name and the names
of the group members whom you met and discussed the paper with, both to
acknowledge their potential influence on your ideas, and so I can make sure that
group discussions are functioning okay. Do not list the names of anyone with whom
you did not discuss the paper. You'll still get credit for the critique, but I might
bug you about what's going on with your group. Here's a template (keep it simple and plaintext):
My name:
I discussed with these people:
Problem (and why important and/or hard and/or why doesn't prior work address it):
Key insights / contribution / strengths:
Weaknesses / fallacies / flaws:
Unsolved problem(s) / opportunities for future work:
Difficulty/ies in understanding:
I'm planning to send your critiques to everyone.
Please read everyone's critiques before class.
I will periodically grade and provide feedback on your critiques.
In-class discussions
Discussion leaders should prepare to lead the class in a discussion of the paper's most interesting points.
Note that the discussion is not a presentation, i.e., it's not like a conference talk. Everyone has already
read, discussed, and critiqued the paper (and read others' critiques).
That said, these papers are hard and will have material that's new to most of us, so as much as half
of the time should be spent leading discussion of challenging points in the paper. The rest
of the time should be spent leading discussion of questions or topics that go beyond just understanding
the paper's key points, such as discussing strengths, weaknesses, connections to related work,
and opportunities/challenges/ideas for future work.
Having slides is not strictly necessary, but slides are often helpful to put up a few diagrams from the paper (or conference talk),
and to display your discussion questions.
It's often a good idea to draw on the whiteboard during a discussion.
You may obtain the authors' slides if you like, but don't use these slides (they will cause you
to give a presentation rather than lead a discussion). It's okay to copy-paste diagrams from the slides to aid
discussions. It's also okay to just copy-paste diagrams from the paper's PDF. Give appropriate credit somewhere
on your slides (e.g., acknowledge all help on the first slide of your slides).
Timeline for group discussions and critiques, and for discussion leaders
Papers
- Finish reading paper and discuss with small group: by morning of day before class (to give enough time to write your critique)
- Write & send critique: by day before class
- Read others' critiques: on day of class
Discussion leaders
- Start reading: at least 10 days before class
- Finish reading paper, outline discussion, and start preparing discussion questions (you might want to contact me at this point): 7 days before class
- Contact me to set up a meeting (include full availability and your current status): 5 days before class
- Send outline (i.e., points & discussion questions) to me: one day before our meeting
- Post slides to Piazza.
Arrive 10 minutes early to class.