Just wrote a short summary of my method to a friend. Might not be terribly readable for people who haven't used Roam, so take it for what it's worth. Hope it's for some use :-)
What I do is an interaction between the following:
Article Tags Hypotheses Synthesis Prediction Research
So I might read an article, then tag the article/figures with the relevant tags. If it's for a specific project, I keep a list of tags on a separate page. This ensure that I use the tags consistently, so that I can find all the relevant articles later.
I purposefully keep the tags general, like "psychiatric conditions" rather than "depression, mania, schizophrenia" etc. When I reach the hypotheses page, this means that I can form relatively general hypotheses, and then nest them into gradually more specific hypothesis that are specifically supported by the articles.
Then, I use those hypotheses to form new predictions. These predictions turn into new research questions, which I either explore via the literature or doing my own experiments.
An example of this might be:
"Wood 2018"
#Procedural Memory #Habits
Habits are part of procedural memory [[H]]
Habits are part of procedural memory [[H]] & Procedural memory is stored in the basal ganglia [[H]]
Damage to the basal ganglia disrupts habits [[H]]
Seger 2011
This affords me a lot of flexibility in total processing time; I don't spend a lot of time forming hypotheses that aren't interesting/necessary for me, I can quickly tag a lot of sources, and I can quickly collate those sources on a specific topic when necessary.
You might say that the first half of the workflow is "induction", the second half "deduction and confirmation".
I would love it if you had any questions or comments, but just writing this up has been useful for me as well.