Month: June 2019

Writing the first technical paper is hard. There are so many things that you want to write, and there are only a few pages that you are allowed to use. Once everything is put together, the manuscript does not read like any other paper that you have read. What went wrong?

Helping someone to write his/her first paper is also hard. I had my fair share of struggle with first time writers, whether English was their first language or not. The good news is that paper writing is not rocket science, everyone would eventually get it, sooner or later. To make this process a little less painful, mostly on my side :O), here is a checklist for you to use to avoid some commonly made mistakes:

1. Not following the standard technical paper structure. There is no real need to innovate on the structure of the paper. Typically, we follow this order for a robotics paper: introduction/related work/contribution statement, problem statement, algorithm, experimental setup, results/discussion, conclusion and future work. The “related work” can be presented in the introduction section, as a separate section after the introduction, or near the end of the paper (seems to be a new fashion);
2. Not paying attention to logic. Make sure to carefully design the paper before start writing. Every paper needs to have its overall purpose, theme, and flow. Every paragraph or sentence also needs to be logically connected to the ones before and after it;
3. Not giving a clear and compelling argument at the very beginning of the paper.  Clearly identifying the objective in the first few sentences would reduce guesswork for the reviewers and channel their thinking to be aligned to what you want them to think. Pointing out the research gap and the potential impact would allow reviewers better appreciate the presented work;
4. Not clearly framing the work in the literature. The main purpose behind the literature review section are: 1) to find the gaps/needs in the previous works to help identify the contribution of this paper; 2) to show that this work is built upon understanding of the history and the state-of-the-art (instead of coming out of thin air);
5. Not clearly pointing out the innovation/contributions. Reviewers would be looking for this to make a decision, so try to help them out!
6. Mixing problems with solutions in the problem statement section. This can have two negative effects: 1) it causes confusion and makes it harder to present the actual solution later; 2) it reduces the possible solutions to the problem and limits the imagination of the reviewer while reading the paper;
7. Giving away the answer too early/easily. This is related to the previous point. Most of the time, if you present a solution right after stating the problem, people would trivialize it. This is like a movie spoiler. Present the problem, point out the challenges, let the reader fully appreciate the magnitude of the problem, then enlighten them. Writing a paper needs some serious storytelling skill!
8. Not detailed enough so that the reader can reproduce the research or/and not concise enough so that the reader won’t get bored;
9. The reasoning/argument is not bulletproof. Every statement needs to be backed up by facts and sound logic;
10. Not providing proper evaluation of the results (e.g., metrics, statistics) and not providing confidence that this is a reproducible result. Spend some time to learn how to design experiments properly;
11. Not providing insightful discussions of the results. Don’t just state the obvious!
12. Not properly recognizing the limitations of the research. Every work has limitations. Typically this can be discussed in the assumptions (in the problem statement section) and in the future work section at the end;
13. Not providing enough diagrams in the manuscript. A picture sometimes worth a thousand words. Try to have at least one figure, diagram, or table in each page. Attaching a cool video with the paper would also be important (especially for a robotics paper);
14. Not fully describing every symbol used in equations;
15. Not following the required format specified by the publisher;
16. Having typos, grammatical errors, and formatting inconsistency;
17. Forgetting to acknowledge sponsors, donors, other non-coauthor contributors.

Note that one of the most important parts of any form of communication is to understand the audience. For the papers, the reviewers are your peers but not necessarily very familiar with the specific problem that you are addressing. Their job is to make a binary classification (often) using the shortest amount of time (sort of like speed dating). Our job is to fight a series of cognitive biases of the reviewers. For example, it’s not uncommon for a reviewer to make a quick judgement after reading only the first few paragraphs (decisions based on incomplete information and heuristics). Once this judgement if formed, it’s difficult to change it. This is because they would be (subconsciously) looking for evidence to reinforce what they believed (confirmation bias). Now you see why the first sentence, the first paragraph, and the first page are so crucial to the success of a paper. Another human bias (availability heuristic) is that people value the most recent information more. This means you need to remind the reviewers all the good stuff about your paper at the end of it…

Well, so far we only talked about the organization aspect of paper writing. How to make sure what you published is not junk (instead, something that would make a technical impact to the community)? That’s another important topic, which we shall discuss at another time.