The Startup of a Lean Doctorate

Using agile practices to tacle a long-term research project

 19 October 2018  |   15 November 2018
  phd tools agile

At the start of my doctoral study, things are a bit woozy. The use of that word “bit” may be the inverse of exaggeration: shaping an abstract idea into a malleable and well-defined project is a process that can easily take up months. After that you still have to “start” actually doing stuff in context of your invented project! Most doctorates are classical examples of a waterfall process: come up with an abstract, do your (and a lot of) literature study, reach certain milestones. It’s clearly visible in the Arenberg Doctoral School Roadmap.

As an agile software engineer, that uncertainty leaves me with an unpleasant feeling as I struggle to define clear goals on a weekly basis. Sprints of 2 or 3 weeks are out of the question… Are they? Why should they be? A quick “agile doctorate” lookup in Google Scholar nets me academic papers like An Agile Approach to the Doctoral Dissertation Process and The Conclusion. The only problem with these papers is that they propose to change the formula all together in an experimental context. There is - surprisingly - next to nothing academically published about someone’s PhD process in an iterative or agile way!

How to be an Agile Academic

Maybe we should ask Jeff Sutherland on how to achieve this, but I doubt the SCRUM Alliance principles were in his head when he did his doctorate. Katy Peplin wrote a nice article about being an agile academic that advocates for principles like getting feedback, fast(er) and using Test Driven Design. As some wise men once said,

Fail fast, fail often.

Early feedback

To fail means to get feedback. To get feedback means to publish. That’s right, publish as early and as often as possible! But how can you publish papers when you haven’t even started researching, let alone writing something useful yet? I’m not only talking about lean publishing of academic papers (as Jake VanderPlas said in defense of extreme openness), but also about concepts. Anything you put on (virtual) paper can be used to get feedback.

For example, Katy identified daily stand ups used in agile teams as a useful way to identify what was blocking her progress by doing a stand-up with… yourself! That way she got feedback from herself. A PhD can be a lonely process indeed, but it doesn’t have to be that way. You can change that yourself by explicitly asking feedback to your peers, colleagues and friends.

My way to get feedback is to follow in Jake’s footsteps and to simply publish everything on Github. Add a big “feedback” button that redirects to the Github Issues page to invite people who come across your work to leave valuable insights. Why only publishing a paper, if you can publish everything related to your work, including papers? It contains reports of meetings, brainstorms, concept items I’m working on, definitions, bibliography, … Literally everything. Document everything - a useful tip!

Another way to get feedback if you’re developing is using your trusty Unit Tests and Continuous Integration system. As I’m not coding for my PhD, this might not apply, you might think. But actually, I can still take those concepts and try to apply them to my writing:

Publishing these pages as a website using Hugo makes it very easy to share a link and ask someone for feedback. Of course, before simply pushing everything into the open, your supervisor (and possibly your Faculty) should agree with this. But after showing the advantages of being open, I’m sure that won’t be a big problem.

Demo or die

Closing the feedback loop might not be that simple: professors are usually extremely busy and have a tight schedule, so don’t expect them to read every long piece you’ve written. Most PhD students meet with their supervisor once a month, just like me. I know students who are less lucky and have more trouble squeezing in a meeting. If you’re really unlucky, your supervisor might not show any interest at all in your research. That’s quite sad and should not happen.

In most agile software development companies, teams present their progress weekly or every two weeks. This is a short moment of reflection that allows for (in their case last-minute) feedback from all stakeholders. I try to emulate this principle by meeting with my supervisor once a month, and meeting with another interested party in the field - also once a month. If you plan this in alternating two weeks, you can gather feedback every 14 days.

The added benefit of gathering feedback from people in the field is another important angle of approach. The result is not a pure theoretical academic work but a good balance between theory and practice. Our Engineering & Technology Faculty at KU Leuven obliges PhD students to write a valorisation report to prove applicability. Another thing more easily checked off…

Reading

Another way to tap into the brains of others is to simply read. Reading books related to your subject will spark new ideas and might even make you rethink certain options. Static feedback is still feedback, even if it’s your own responsibility to interpret written opinions of others.

I am not talking about your typical academic literature, but also about popular science books, philosophy, sociology and psychology, … - anything that even remotely touches upon your research questions might be worth looking into. I prioritize my reading list based on interest instead of relevance: embrace the possibility for new ideas coming from any direction. For example, I’m researching non-cognitive skills in software engineering; Work Rules!, Mindset and The Reflective Practitioner are currently on my list.

Planning

Another useful tip you hear often might be to “try to work on a weekly basis”. They are talking about planning your week. If planning a whole doctorate is too difficult (it is, and it’s scope will change often) - simply plan each week. I use Trello for that and you guessed it - it’s open to the public. It’s a bit more intricate than your average “todo - doing - done” list, but not by much:

  1. Todo - backlog: anything in here. Things I will need to do in the next months - years. Very vague items. Might and probably will be split in multiple items.
  2. Todo - Week X [Deadline Y]: Something to work towards, for example a seminar you’re attending or holding next month, or a brainstorm week you’re holding with a research group where you can get feedback from your supervisor (or others).
  3. Todo - Week X: the current week. What will you be doing this week? Try to be as descriptive and concrete as possible.
  4. Doing
  5. Done: timestamped.

Apart from that, I’m relying on the following labels:

The lean Tools used to battle with

1. Use analog tools

I have never had that much freedom in any long term project I’ve ever worked on, than my own doctorate. I am the one who decides what to focus on and I am the one who decides which direction to go to - of course all in dialogue with my supervisors. But that liberating feeling generates a lot of wild new ideas. Those ideas present themselves on unexpected moments - so you better make sure you have something to write with and something to write on. My love for pen & paper is well known. I jot down ideas, mindmaps and summaries of papers and manage to fill at least 2 A4 papers every single day.

As with all things, if you don’t re-read your notes, they will be lost. See my post about journaling in practice for details on how to set up an easy, working system.
After the rough sketching skep, ripened and mature ideas might be worthy of a more substantial body.

2. Use simple digital tools

Writing requires a… typewriter? Text editor? Sublime Text. Pushing to a github repository means we’ll be using git and the command line (iTerm) a lot. I’m a big fan of Markdown thanks to small but great generators like Hugo used on this site. Instead of redirecting everyone to the Github repository homepage, I converted my PhD repository to a Hugo website and published it as a Github Page. That means written Markdown files are instantly and easily readable by a bigger audience and sharing the work through a website is a bit easier.

Sublime works well in combination with Git and Markdown; especially with plugins like WordCount, MarkdownEditing, GitGutter, BracketHighlighter and Compare Side-By-Side. My choice for using Markdown instead of LaTeX might sound strange for academics, but Hugo’s publishing skills are simply unmatched if you want feedback, fast. Things like Pandoc and AcademicMarkdown can be used to convert .md files to academic .pdf papers, like Eric J. Ma did. More on that later when I’m on the verge of actually publishing something in an academic journal.

Anything else?

I’m sure my methods of bootstrapping my research can be further enhanced by people who’ve been through the whole process and are also agile advocates. If you think my work might benefit from some other tool or practice not mentioned here, please let me know by adding an issue in the Github repo as feedback. Every remark is greatly appreciated!

In the meantime, take a second to skim through my work so far. Thank you!

Productivity Tools on all platforms

Matching my most frequently used tools on OSX, Linux and Windows  28 August 2018

Computer Science learning pathways

Categorizing essential Computer Science knowledge  29 June 2018

Reverse engineering a curriculum

Can a unified software engineering philosopy shape the right context in a curriculum?  15 June 2018

 Top