HW1: Reaction paper

Due date: October 24, 5pm (no late submissions allowed)

The idea of reaction papers is for you to familiarize yourselves more in depth with some of the material covered in class, and do some reading beyond what was covered. At the same time, you should be thinking beyond what you just read, and not just take other people's work for granted.

Consequently, your reaction paper should cover at least two research papers on closely related topics. The papers should be related to topics covered in the first 6 weeks of classes, so broadly robustness and fairness. At least one of these should be a paper that was not covered in class, either in the lectures or in the class presentations . In other words, go find something that I may not have heard about yet. One of the two papers can be a paper covered in the class presentations (for e.g. you could continue from a mini-HW submission you felt excited about), but it should not be a paper you yourself presented in class . You are welcome to discuss more than two papers, if you feel that it fits well with your narrative; in fact, it is probably a good idea to discuss several additional related papers at least briefly to contextualize the papers you are focusing on.

The reaction paper should be individual work. In particular, you should go look for the papers yourself, and not just ask others about papers they discussed and use the same ones. You may use AI to help you understand papers you choose, but the reaction paper itself should be your own work.

Your writeup should be about 3 to 5 pages with normal fonts and margins and single spacing. If you have a lot to say, in particular along the lines of new ideas, it can be longer. If it is much shorter, you probably are not going into enough depth. The organization should be roughly as follows.

• About one page should be used to summarize the main content and results of the papers you are discussing. How do they fit in the field, and what you have learned in class so far? What is the connection between the papers you are discussing?
• About one page should be a judgment of what is in the paper. What struck you as particularly interesting? What were the authors missing? Was anything particularly unrealistic? This section should go into a bit of depth, for e.g. typically more than in mini-HWs.
• About one page should be a discussion of what you feel may be an interesting step to take beyond what the papers are doing. Perhaps you have an idea of a better model for something? A better algorithm? A better evaluation? Perhaps the papers suggest techniques that you would like to apply elsewhere? Obviously, your ideas here will not be completely worked out. They should be brainstorming ideas. If you have the time to look up some more relevant literature, or can cite it off the top of your head, all the better. Just as a measure of calibration, this should be the result of thinking about possible extensions for a couple of hours or so. Not a 5-minute output, nor a careful research project (yet).
• Obviously, these parts need not be separated into sections. You can combine them in any way that makes for a good narrative.
  
  

The paper will be due on Gradescope. Please note that no late submissions are allowed, so plan accordingly.

(Thanks to David Kempe for letting us borrow his idea of a reaction paper.)