Great discussion here in response to the COW team's queries!
I think @johnsarmi's suggestion of starting in a bottom up way with identifying outcomes is really appropriate at this stage of the development of COW as a set of materials. To that end, observing classrooms is a key way to figure out what students might be learning. One does need to move fairly quickly into thinking: What are these constructs I am seeing? What connections do they have to things documented in the literature? This will help with moving from grounded theory-driven approaches into measurement. The two approaches really complement each other nicely.
@annaosipova raises some important points about challenges associated with managing classrooms in PBL. This in my view falls into a "known issue," meaning we can anticipate variation in management of PBL, and so we should be thinking from the start about how to design for it. And evaluators should be attuned to looking at how people manage the workflow.
@jfrickey's suggestion of a meeting to look at implementation evidence together is a great one. We've done that at various points in my own DBIR work. Folks can see resources related to developing and using implementation evidence at the LearnDBIR web site.
I have used reflection journals of the kind that @juliadaniel suggests might be useful. I think they could in fact be helpful here. A key here is to give some prompts that elicit responses that allow you to draw the inferences you want to about particular things. Specifically, if you want to know about SEL growth, you may need to ask some open-ended prompts about ways that students collaborated or reflected that day, or ways that teachers sought to support students' developing empathic responses toward characters they are reading about.
Joe asks about making practical measures practical. For one, you have to leverage technology in the way the team is doing. Building measures in Google Forms is a simple and effective way to do this. Forms can produce graphical summaries with little data transformation required. The trick is when you want to protect teachers' privacy but then look across teachers at patterns of data. Then, you have to do some data aggregation, which can be time-intensive. At present, we don't have the perfect infrastructure for practical measures, but I think that is key. I do think @jfrickey's idea of a common meeting time would help -- it creates a kind of deadline for the team to get data analyzed. I've experienced this very positively on projects I've been involved in, where a PD team asked to see the implementation evidence the PD researchers were developing, so they could think about how to improve their work. A presentation creates a real sense of urgency for the research team to get to the data analysis they might otherwise put aside.