I found an article titled "a routine for knowledge discovery within a research group" (https://graphthinking.blogspot.com/2021/01/a-routine-for-knowledge-discovery.html). Here are my thoughts on it. NOTE: this article may be "live" and updated occasionally with no warning and no revision history (aside from what's in git).
My initial reaction while reading the beginning of this article is that the scientific method is not translatable to social interaction or collaboration because those latter two things are difficult in a fundamentally different way. In particular, I suspect that, among other things, collaboration is much more heterogeneous than the scientific method. The more heterogeneous a problem is, the less useful general solutions are ("general" being a truly agnostic solution and not a covering of individual specific solutions). Moreover, I suspect that most attempted solutions to this particular problem (if not the superclass of heterogeneous problems in general) will only address a common, thin "substrate" or sub-problem. For instance, you could attempt to address the collaboration problem by requiring all communication to either go directly through a textual medium (email, chat, or blog) or to be mirrored to text later. This doesn't actually solve the collaboration problem at all - it solves a few tiny sub-problems of communication (knowledge not being captured, people mis-remembering things, etc), but will have absolutely no effect when the team members don't initiate communication in the first place because they're busy working on their own section of the project in isolation from the others.
"When: every two to three weeks, an artifact should be produced and assessment of progress made. Assessment involves the owner of the experiment and peers on the team and the team manager." I disagree with the timing schedule. Artifact production on a two-to-three-week schedule, or almost any fixed schedule on a period from days to months, seems to be mainly geared toward providing progress reports to a manager, and not conducive to actual research. Artifacts are generated from experiments naturally. Instead of intentionally only capturing them when it's time for a "progress report", they should be captured at generation time, made accessible to the team at the same point, and any progress reports should merely point to interesting or notable (but already existing and available) artifacts.