Or How to Change the World with Six Pages?
Writing scientific papers is not a trivial task. The fine art of explaining what you have done requires specific training. This training is often missing, and learning through trials is not feasible. I am reporting a how-to guide that my colleague Sebastian Wallkötter taught me on a rainy weekend in Paris. This guide easily applies to computer science/engineering papers. The experiment section mentions the APA guidelines.
Three main take-away messages:
- There is no way to convince someone else that your work has value if you don’t believe it has.
- It is highly probable that you are not making a revolutionary contribution, but this should not stop you to keep pushing on your research (“Your contribution is one step towards changing the world” cit. Wallkötter).
- Simplicity and shortness are quality factors.
Before going to all the sections, take a marker and stand in front of a whiteboard or a notebook. Write down the answer to this simple question: how is your work contributing to the current state of the art?
- Context: What are you speaking about? (one sentence or two)
- Research: What did you investigate?
- Evaluation: How did you evaluate your research?
- Results: What are your main findings?
- Take home message: What is the impact of your work?
- Motivation: Why is the topic important?
- Research Context: What has been done?
- Gap: What research is missing?
- Contribution: How is your work solving the gap?
- Impact: What are the benefits of your work?
- Literature Review: Are there literature reviews about the topic? What are they investigating? Extract keywords from them.
- Low-level Papers (e.g., Experimental): What are the relevant papers that you find with your query?
- Gap analysis: What is missing in the state of the art? Show that your work addresses a gap in the current knowledge?
- Commonalities and Differences: What are the commonalities and differences between your work and the state of the art? (be specific about your contribution)
- Position: Where do you position your work? How is your work contributing to the state of the art?
Never talk badly about others’ work, those “others” will be your next reviewers. Show that you know about the topic. It established a certain amount of authority in the field. Have a look at “How to related work” for further information.
- Theoretical formulation: What algorithm are you using? How do you formalize your problem?
- System engineering: How do you implement your algorithm? What are the components of your system?
Here make clear what is your technical and theoretical contributions.
- Hypothesis: What are your hypotheses? What are your claims about a casual relationship in the world? Are the hypotheses that you claim falsifiable? (make sure that you can test your hypotheses, e.g., you cannot measure if there is an undetectable dragon in your room)
- Design: What are your conditions? What independent variable are you manipulating? Which dependent variables are you measuring (specify how those measures will help in exploring your hypotheses)? Did you use standardized questionnaires (cite and share them)? Did you do an experiment with humans? (ethical clearance)
- Participants (if you have): What are your participants (e.g., adults? children?)? How many are they? What are their demographics (e.g., age, gender)? How are them distributed among conditions? (how many per condition? how many females and males?)
- Materials: What do people need to replicate your experiment? What did you use for your experiment? Computer (if you work in simulations report CPU)? Robot? Tablet? Room/Environment? Language?
- Procedure: Did you randomize your subjects (humans or agents)? How was the experiment conducted step by step? How long did the experiment take?
- Results: What did you measure following the order of your hypotheses? Which type of analysis did you run? Report your results consistently with respect to the type of analysis. Do you have any other exploratory results to report? Show the chart of the significant results that you obtain reporting the p-value.
See APA guidelines for more details.
- Technical contribution: Why your contribution is valid and real? (address potential arguments that could be made against your contribution) Did your system work as expected? What are the peculiarities of your system?
- Hypotheses discussion: Were your hypothesis confirmed or rejected? Why do you think they were confirmed/rejected?
- Limitations: What are the limitation of your system? How do you plan to address them?
The main idea of the discussion is defending your contribution. Formulate your discussion as a debate, give arguments and counterfactual arguments.
- Research: What did you explore?
- Methods: How did you explore it?
- Evaluation: How did you validate your approach?
- Results: What are the results of your experiment (with respect to your hypotheses)
- Contribution: How did your work contribute to the state of the art?
- Take home message: How did your work impact the world?