howtowriteapaper

How to write a paper Thomas Ågotnes Or... • Some tips for beginning students on how to become good at writing formal logic papers, in English, that are accepted for good conferences or journals Why? • If you want to have an academic career, you have to publish • Writing is the best way to think about things • Writing is research Plan • Review criteria - what is a good paper? • What to do and where to start • How to structure a logic paper • The submission and reviewing processes • Publication outlets Typical review criteria 1.Relevance 2.Originality and novelty 3.Significance 4.Readability and organisation 5.Technical quality and soundness The importance of results • For a paper to be accepted, you need results • A good idea is not enough • There must be some kind of evaluation of the idea • Analytical (most common in logic), or • Empirical • In logic papers results are most often (non-trivial) theorems or lemmas • The paper cannot be just a list of definitions Significance • Significance = good idea + good results • It is easier to get a paper with a mediocre idea but strong results accepted, than a paper with a great idea but weak results Presentation • Presentation, both the quality of the technical details and the quality of the English text, is very important • This is in particular true for papers about formal logic • Precision is extremely important! • Don’t underestimate the importance of presentation • If the reviewer is confused and don’t understand exactly what you mean, the paper will be rejected • You have to become a good writer - in mathematics and in English! How to become a good writer? • Read! • you should read a lot of papers in your area • pay attention to • structure (sections: Introduction, Background, ...) • precise definitions • how proofs are presented • future work and open problems! How to become a good writer • To be a good English writer, read as much as possible of anything in English • books • newspapers • magazines • Watch movies and TV series without subtitles • Watch CCTV • Learn to start thinking about your work in English Tools • Learn LaTeX! • Takes some time to learn, but after you have learned it everything becomes much easier • Find a good text editor, like Emacs • Use a spell checker! The writing process • Start with writing down technical details (definitions, etc.) early • While you are working on the results • extremely important that definitions are precise • When you think you have a proof of a result on a piece of paper or the white board, write up the statement of the result and the proof in LaTeX • you will probably find mistakes or holes • Write all proof details in LaTeX. You can remove them or put them in an appendix later The writing process • Write - read - write - read - ... • very important to read your own paper very carefully, many times, while you are writing it • print it on paper! • Write the text after you have finished with the main technical results Originality and novelty, and citations • You have to argue that you do something new, something no-one has done before, otherwise your paper will be rejected • It is therefore very important to discuss all related work • It is your responsibility to convince the reviewers that it is new, it is not their job to convince you that it is not • Reviewers can be crossed if you don’t cite them! • Also cite yourself, even if the submission is double-blind • But don’t cite your own papers disproportionally Tip • In the beginning try to write a “boring” paper, a paper with a “small” idea • Maybe just a small extension or variant of what someone else did • But do it very carefully • To learn the craft • Don’t try to do too much before you know how to write a paper What to do? Where to start? • In the beginning, it is much better to try to do something “small”, but to do it as properly as possible • The first papers you write are your training in how to do it • the most important thing is not to do some very important research, but to learn the craft of paper writing • Good source for ideas: “Discussion” section in papers, on future work One general approach One general approach 1. Find some class of mathematical structures to use as models • Existing models from logic (epistemic models, ..) • Existing structures from some other field, such as models of games from game theory • Extend existing models in some way (from single- to multi-agent, add epistemic relations, ...) • Define a new type of models One general approach 1. Find some class of mathematical structures to use as models • Existing models from logic (epistemic models, ..) • Existing structures from some other field, such as models of games from game theory • Extend existing models in some way (from single- to multi-agent, add epistemic relations, ...) • Define a new type of models 2. Find some interesting properties a model can have or not have • common knowledge of some proposition, existence of Nash equilibria, .. One general approach 1. Find some class of mathematical structures to use as models • Existing models from logic (epistemic models, ..) • Existing structures from some other field, such as models of games from game theory • Extend existing models in some way (from single- to multi-agent, add epistemic relations, ...) • Define a new type of models 2. Find some interesting properties a model can have or not have • common knowledge of some proposition, existence of Nash equilibria, .. 3. Try to construct a modal language that can express those properties One general approach 1. Find some class of mathematical structures to use as models • Existing models from logic (epistemic models, ..) • Existing structures from some other field, such as models of games from game theory • Extend existing models in some way (from single- to multi-agent, add epistemic relations, ...) • Define a new type of models 2. Find some interesting properties a model can have or not have • common knowledge of some proposition, existence of Nash equilibria, .. 3. Try to construct a modal language that can express those properties 4. Investigate meta-logical properties of the resulting logic Expressing properties: the local view M, s |= ' model some kind of additional context related to the model epistemic model: state game model: e.g., outcome property Typical logic paper: structure • The title • Don’t promise too much! • The abstract • Not too long • Just one or two sentences describing the problem and the motivation, and then a description of the main contributions of the paper • good English language here is very important • Reviewers are often assigned to your paper based on the title and abstract, so it is a good idea to use words/terms that will attract the right reviewers Typical logic paper: structure • 1. Introduction • what the idea is and why it is interesting/important • argue that nobody did it before • overview of your paper • good English language here is very important Typical logic paper: structure • 2. Background • definitions/results from the existing literature that you are going to use • important to make it very clear what is from other papers, and what is your own contribution Typical logic paper: structure • 3. Definition of your logic • language • formal semantics • examples • can express interesting/useful properties Typical logic paper: structure • 4 - ??: Meta-logical results • Validities • Axiomatisation and completeness proof • Characterisation of expressive power • Decidability proof • Characterisation of computational complexity • Relationships to other logics • ... Typical logic paper: structure • x. Discussion • Summary of your contribution • More discussion about the relationship to other work • Suggestions for future work Submitting papers • Read the submission instructions in the call for papers very carefully • Your paper will be rejected if you don’t follow the instructions wrt. • double-blind reviewing (remove your name from the paper) • the number of pages including appendix, bibliography, etc • very important • trick: you can give an URL to where the reviewers can find additional information such as long proofs, etc. • deadline The review process • Typically you will receive 2-4 reviews • They can range from very helpful, to useless • The process is imperfect: some good papers get rejected and some bad papers get accepted • Rejection doesn’t necessarily mean that your paper is not good • But you should read carefully what the reviewers say, and revise your paper before you submit it again somewhere else The review process • Sometimes there is a author feedback phase where authors can respond to the reviews before the final decision is made • Be sure to be very polite in your response • You should only try to clear up misunderstandings • Don’t be argumentative Reviewing • If you get the opportunity, review as much as possible • You learn a lot from reading papers very carefully Publication outlets • Workshop • Review process but usually no proper publication • Sometimes accept papers with good ideas but weak or no results • Can be useful in the beginning of a paper • Conference • Good conferences have a thorough review process and proper publications • Page limit • Special issue of journal • Often good reviewers • Page limit • Can be easier to get accepted than normal journal submission • Normal journal submission • Often no page limit • Can be unlucky with reviewers Publication outlets • Every outlet has its own community and its own “style” • read! • Don’t submit to several outlets at the same time • An extended revision of a conference paper can be published in a journal • some kind of significant extension • more results, proofs, ... Conclusions • Writing in general, and writing academic formal logic papers in particular, is something you can learn • You have to put in the time • You will have many papers rejected, no need to worry about it • Conferences typically 15-25% acceptance rate • Again: read a lot, write a lot