updated: Mar 2, 2016

length: 13 min

Research & Scientific Writing Process

How to research effectively and systematically with Evernote and write scientific papers from digital notecards.

Research & Scientific Writing Process

While the human brain excells at establishing connections between different ideas, it is not very good at remembering specifics (especially my brain) — quite the inverse of the capabilities of computers. Niklas Luhman with his Zettelkasten I had to find ways to cope with this while writing my second Bachelor Thesis about Donald Davidson’s philosophy of language and his critique of the Turing Test. Research in Philosophy, especially, requires a lot of reading, doesn’t have an inherent structure, and the goal is often not very clear from the beginning. So a system to capture the read information is crucial to make use of this information effectively. But really in any longer research project or career, knowledge management certainly is essential. Furthermore, before you have written something down, thoughts are rarely ever as structured as they need to be to really understand something. I have long been a fan of the Zettelkasten (file-card box) method as developed by German sociologist and system theoretician Niklas Luhmann. The main idea of this method is to capture all information in notes which are organized primarily through links between them — giving rise to a system which enables you to communicate with your former thoughts. Luhmann describes it as follows:

A Zettelkasten can be recommended as a partner for communication because of the technical-economical problems of scientific work. Without writing, one cannot think — at least not in a sophisticated and connected manner. …But if one has to write anyway, it is prudent to also use this activity to create a competent partner for communication through a system of notes. Niklas Luhmann Communication with Zettelkästen, original quote in German

In this post I want to give you some insights into how this idea has inspired my research and writing workflow using Evernote and gingko.

I do nearly all of my reading for research digitally, mainly because it enables me to always carry everything with me and it allows me to find things faster, thanks to the amazing power of Spotlight in Mac OSX and search in Evernote. The most important step is to not only read and maybe highlight some passages, but to take notes on all important things you read. For complicated texts (like some philosophical papers) this requires several steps. Maybe you need to read the text once and highlight passages (I’d suggest to use at least one color for key areas and another for examples) and then read it again and take your notes. I have often encountered that until I try to summarize the text and write down some of my own thoughts on it, I may think I have understood it, but I actually haven’t — the more complex and abstract the subject, the more this applies. Also in the note taking process I sometimes take a two step approach, especially for very complex and dense texts. First I try to reconstruct the arguments in my own words, then I write my own abstract based on the notes I have taken previously. In that way I can condense my knowledge and it becomes accessible in different layers of detail.

My reading setup, text on the right, notes on the left.

Why Evernote?

So where do I put those notes and how do I organize them? A common approach is to organize notes hierarchically into categories or on a project basis. The file system of a computer enforces such a hierarchical structure. In my experience, however, some of the most important connections are not captured in a hierarchical structure but emerge when we make connections which we haven’t anticipated. The structure usually changes with the question we have in mind. Therefore while a hierarchical system is very usefull for structuring the writing and presentation process, it cannot act as a general knowledge repository for your research process. You might say that it is good to enforce the hierarchical structure because the presentation is your final goal, but then you assume that the knoweldge is not relevant for future work and that you can already know the structure of your final output before you finished your research. Clearly you shouldn’t assume either: (1) any development is based on the fact that old knowledge can become relevant again and (2) you should separate your writing and research process to open yourself to truly learning about your subject of research and be open to being corrected about your assumptions.

While there are other tools like the Zettelkasten App, my preferred tool by far is Evernote. Evernote’s slogan is “Remember Everything” and actually Luhmann described his Zettelkasten as a secondary memory too. Not only does Evernote have apps for all operating systems, have a beautiful interface, back up your notes with version control in the cloud, allow you to easily share them, and make it easy to save websites as notes, it also makes it easy to link notes to one another, tag them, and filter them. Most importantly it allows you to attach any file to a note and has an extremely powerful and blazing fast search function which even searches inside attached files and performs text recognition on images. Therefore it is seriously able to replace the normal file system. There are not only a lot of notes in my Evernote, but also a lot of files (especially research papers) which I have only stored in my Evernote and nowhere else.

The great Search function and the clear card display in Evernote make it easy to find relevant notes and recognize them visually. Links provide a second level of organization allowing you to create wiki-like structures.

There are a couple ‘gotchas’ about how to set up Evernote best. I highly recommend integrating all general items into one notebook and not start putting things into different categories. The only exception for me are notebooks that need to be shared with others and one notebook for official documents like bills, passport scans, and contracts. For all general knowledge, it’s simply impossible to attach a single category to it and search is the primary way to find my own notes anyway. To structure notes, I use tags and links. Tags replace categories and folders since it is easy to filter for all notes with a specific tag, but it also enables me to put each note into many categories and filter for arbitrary combinations. Links are great for creating structures and enabling the subdivision of information into different notes while keeping them linked together. For example: I have inline links if notes are referenced in other notes, and I have notes with links to all the relevant articles from a certain book, or on a certain topic, or for notes from different sessions of one seminar and the corresponding reading notes on the discussed papers. An important part of making the most of such index notes is to add links from each of the individual notes back to the index, in that way wiki-like structures emerge. For certain projects I have created full-fledged wiki systems with several layers of indexes in that way. It is a bit of work to tag notes properly and even more so to link them together, therefore, I only create links for valuable notes. The value of a note is usually measured very accurately by how much work went into creating it.

The Structuring and Writing Process

While I know some people who like to use Evernote for writing as well, that’s not my cup of tea. I think it is important to take a step back from your notes to create structure and in my experience nothing is more efficient than spreading out notecards or using sticky notes to structure your thoughts on a topic. Physical tools like this enable a much faster way to visualize and organize content. Sometimes I write the titles of corresponding evernotes on the back of notecards to create a link to the digital world. Similarly, I sometimes digitize my notecard layouts and handwritten notes with the great Scannable App. But more often than not, the brainstorming with notecards is just an intermediary step which I don’t need to save anyway. For my thesis, for example, I created the outline by arranging notecards with information about all different concepts in Davidson’s Philosophy that I had read about.

No digital tool can replace the ease and rapidity of physical notecards nor provide the same overview you can get by spreading out physical notecards on a large surface.

After the structuring comes the process of writing or creating some other kind of presentation. For writing I usually like to use Markdown, especially when it comes to research where I want to separate content and representation as much as possible. For writing my thesis I discovered the great gingko app, which structures your text as a tree of notes. This makes it easy to visualize the structure of your paper, keeping things clear, and allows you to switch back and forth between different sections seamlessly. I set up my outline of sections and subsections in the tree and started by writing down bulletpoints of what I wanted to discuss in each subsection including links to the relevant evernotes. From there I started writing the actual text and then commenting out the bulletpoints at the end. Of course the process is not quite that linear and I often need to go back to doing more research on a certain topics that emerged during the writing and then rewriting some sections based on the new understanding. I think this iterative process is crucial to develop a deep understanding, enable a creative process, and create good writing. Gingko enables this iterative process by making it easy to create new versions and to restructure the text by simply dragging and dropping cards around.

Gingko has a great interface that visualizes the structure of your writing and allows you to jump between different sections quickly.

The writing in gingko was very enjoyable and helped me to stay sane during the sometimes tiring writing process.

Exporting the Writing to Different Formats

But it doesn’t quite end with the writing, because the last step is to format. For scientific papers the output is still mostly PDF, but sometimes it also requires HTML for sharing online, or some Office format that has a suggesting or track changes mode that enables much better feedback loops. I know one magic tool which allows you to do all of this at once: Pandoc. Pandoc enables you to convert between Markdown, HTML, PDF via LaTeX, Microsoft Word (and therefore also Google Docs which I prefer), and more. It supports it’s own extremely powerful Markdown flavor which integrates with most LaTeX functions and it is also possible to write custom LaTeX templates and compile with XeTeX in order to use custom typefaces. I favor Markdown over LaTeX because it is much more readable and naturally compiles to HTML as well. For references, I of course use BibTeX, which is also supported by Pandoc-flavored Markdown. I actually tried out several BibTeX managers (including Papers, Mendeley, and Zotero) but settled for writing my BibTeX manually in a text editor because none of them supported sources with both an original and a cited publication and in Philosophy you really don’t want to cite either the original Latin “Descartes (1637)” nor anachronistically just cite the translation “Descartes (1993)”, but both “Descartes ([1637] 1993)”. Maybe creating a simple BibTeX editor would be a great project for the future.

The final pdf generated with Pandoc with BibTeX and XeTeX for custom typefaces.

So that’s it: my whole workflow from research to print.

APPENDIX: Markdown Examples

Here is a litte Markdown example from my thesis:

### Interpretability as the essence of Intelligence

What we have discussed so far depends on accepting Turing's
first claim that communication abilities are representative
of intellectual abilities. As mentioned in the introduction,
Davidson agrees with Turing on this point, since he "sees our
nature as linguistic beings as the key to the possibility of
thought" [@Lepore:2009d, p. 1], a belief springing from his
metaphysics[^metaphysics] and epistemology. (I examine the
latter in more detail in \autoref{triangulation}.) However,
Davidson emphasizes the interrogator's judgement as the
essential point. Not the computer's ability to *communicate*
is essential but the fact that the interrogator can
*interpret* the computer.

> "Turing was right, in my opinion, in taking as the only
> test for the presence of thought and meaning the interpretive
> powers and abilities of a human interpreter."
> [@Davidson:1990tt, p.86]

Davidson intentionally changes the emphasis here since he is
an Interpretationist --- he holds the ability to be interpreted
as not only a sufficient but a *necessary* condition for
attributing thought, introspection into the working of the mind
for example is *not* a sufficient condition for Davidson.

    Davidson argues that "if we have the semantics of a language
	right, the objects we assign to the expressions of the
	language must exist" [@Davidson:1993mm, p. 40]. Since being
	able to communicate successfully is a sufficient condition
	for getting semantics mostly right acording to Davidson
	(see \autoref{interpretation}) ...

And this is some of the Corresponding BibTeX:

	author = {Davidson, Donald},
	title = {Turing's Test},
	year = {2004},
	booktitle = {Essay Volume 4},
	origdate = {1990},
	origpublisher = {\textit{Modelling the Mind}, edited by
	                 K A Mohyeldin Said. Oxford University Press},
	chapter = {5},
	pages = {77--86},

Which together with the template produces this output:

example output