CC-BY-NC-SA_4.0
- Only my HR Delay Decoding project is on this repo (./neuroimaging/)
- Results from December 23rd in jupyter notebook (./neuroimaging/HRdelay.ipynb)
- These results are wrong, specifically the decoding results, due to a bug in the fixed-effect model leading to some non-independence between model training (feature selection) and testing
- Currently fixing the bug by making the code clearer
I have been whining for years about the flaws of science, or rather of the social system for scientific production. I took a few depressions and a pandemic to finally to finally stop worrying and start working openly. Skipping on the details, I will just say that the open science mindset is so much more confortable than the standard one. I must not lie though, it is hard work to change the workflow. I however strongly beleive the individual efficiency loss is nothing compared to the collective efficiency gain. That is probably what makes the process so exciting to me--"finally ALL my work will turnout at least a little usefull". I could go on and explain in detail how open science can be awesome (it is in my opinion not awesome yet, but getting there). There is plenty of hard and soft litterature on that so I'll let the reader make its own mind. I will instead take the opportinuty here to document my personal process for my transistion toward open science, hoping to fill what seems to be a gap in collective knowledge and respond to the comment I have always myself made: "yeah, open science principals are great, but how can I personally make the transition?".
I was, like many, a very stressed researcher. I welcomed collaboration but at the same time was challenged and inhibited by issues of attribution. I wanted to talk about my work but feared getting scooped. A closely related paper came out and I was excited to read it but felt a pinch of worry that it is "too related". I did my best to organize my own work in my own computer, but I did so while constantly trying new expirmental methods that come with new tools, and with a code literacy that was not taught to me but learned by just coding stuff with my friend Google. The results: a mess on which you cannot do otherwise than to pile more mess. Then you need to produce something like papers and a thesis. You roll up your sleeves and pinch your nose and from the 3-thesis-worth mess of a pile, houray! you pull out the string of a story. The relief is short. You are told to just write it, but your bride wants it deeply anchored in your data and the litterature, but finding the strong pilars in the mushy pile of code, spreadheets, various file formats, so many notes, pieces of text, figures, powerpoint presentations, multiple endnote bobliographies with over 1000 references and 3 one-foot high paper stacks, yeah finding the pilars is hard, and they never feel strong enough. You then just need November to deprive you of the sun to tip you into depression. Luckily I have friends and colleagues to support me, and to tell me everybody feels at least somewhat that way. So why the fuck? I wondered! Do we really need to go through that? Pre-agricultural revolution hunter gatherer societies worked on averaged 25 (maybe 35, can't remember) hours a week. Yes we may have needed to work our asses off to stop dying at 12 from an infected wound, but with the technology of a few more century can't we at least reserve some time to figure out ways to work collectively better? I was sure there was a way, but I was stuck, and angry all the time. This yet another breakdown had some good: I started to learn about open science practices a little more, but mostly talked about. Talked about it quite passionatly, often around beers with friends and colleagues but also with anyone who had the misfortune of getting me started. But I was still far from action, I was instead slowly getting back to my thesis, my precious, slowly pulling it out of my pile of mess the classic way.
Then COVID19 happened. Pushed out of office, there was close to a week of dis(re)organisation before I could put down a solide week of surprisingly good thesis work. That is when it struck me: "I, as I scientist, can and must do something to help during this crisis". I happen have a courageous PhD supervisor. Vision and brain science has little to do with COVID19, yet problem solving, organisation and innovation are general skills a scientist can apply to any problem. In the case of my supervisor, he used those skills and many more up his sleeve to set-up a prized challenge for the design of inexpensive and rapidely deployable mechanical ventilators (those were feared to come in short supply at the epidemic peak). Like most of our lab and at the expense of my thesis, I helped, but while I deeply admired everyone's courage, a nagging feeling crept on me: the challenge set to bar to Health Canada specifications, too high to produce usable solutions on time for the first wave. It is in that context that I stumbled on some obscure blog post about an Italian team and their solution for the very real shortage of respiratory devices hospitals were facing. They designed and 3D-printed an adaptor to connect standard respiratory therapy (RT) equipement to a commercial snorkeling mask and turn it into an alternative battefield solution to support COVID19 patients' oxygenation. There was barely any information about the Italian solution: one diagram, few lines of instructions and the 3D design file to print the adapter, which they had just released (on my birth day). I thought: "Hey, we have a 3D printer at the lab, which is in an hospital where I certainly can find the necessary RT equipement. The very first step for this to potentially save lives here is to find a snorkeling mask and see if we can build the thing and whether it actually works. Damn, this should be done now!". So this is when I started to act big time. Lot's of good stories on this adventure with many people and the very instrumental contribution of Tatiana Ruiz, but to keep this brief, I will just say that to be fast we HAD to open, so we did.
The problem is that we did not know how to work openly: what tool to use, what are the legal repercussions, licensing, patenting, liability... And no time to learn best practices. So we just started to write everything down in easily sharable Google Docs, and that became the ScubPAP Documentation Project. We further refined our own view on how we should work openly, and founded a non-for-profit oragnisation, AltPAP Innovation, to support the open work. I barely touched my thesis during spring 2020. Luckily the first wave passed without the need for our last-resort non-approved solution and I could get back to it during the summer. It was however now hard to go back to the old strategy just make sense of my pile of mess, "just get papers out" and rapidely wrap-up the thesis. I started more seriously exploring the open science ecosystem, with the help of another group of people assembled in the Advisory Committee for the Fair Science Platform.
As of the writing of this file here, midnight on November 16th 2020, the second wave is raging, I have until December to submit my thesis :-S and we are setting up an open lab for the validation and training on new respiration technologies. I have almost litterally dumped online data from two of my thesis chapters, and trying to be a little closer to the FAIR principles for the third chapter with a jupyter notebook in this repo. I also want to take notes of this process of going open, which is now well on its way. I will open another file here specificaly for that documentation purpose, but for now, I am going to bed. Good night