One important dimension of conceptual building blocks for synthesis is "size".
Intuitively, having building blocks that are too "large" or complex would make reassembly into a new whole impractical or impossible. Thus, the most basic requirement for synthesis is having appropriate(ly sized) building blocks to start with.
There are a few ways to think about size and why it matters for synthesis.
A simple benefit is just making it easier to find what you're actually looking for! For example, @ribaupierreExtractingDiscourseElements2017 found that
decomposing ideas into smaller pieces also enables us to connect ideas in richer and more meaningful ways. For example, we learn from creativity theory that Z: Representing ideas in atomic units facilitates conceptual combination. This is because Z: Synthesis is a creative act, and Z: Conceptual combination is-fundamental-to creative knowledge production.
It is important to distinguish Atomicity from compression: in creative thought, it is less about Decomposition in the atomic and disconnected sense, and more about flexibly using compression to move between different levels and states of "granularity".
We might wonder: if we break complex documents down in a synthesis infrastructure, what should the component parts look like? What defines an "idea" level, or an appropriately "small" building block for scholarly synthesis?
This should be nuanced, though, because restricting ourselves to only linguistic / symbolic representations would probably be a mistake. There are other forms of knowledge, visual and otherwise, that are important "units" that can't be reduced to their "underlying" linguistic representation.
Scholars encounter a problem with "chunks" (papers, sources, etc.), either ones they have already, or ones they seek out.
They want to construct a new understanding. To do that, they need to decompose the chunk of the paper/source/chapter, etc. into the component parts they care about, to be able to manipulate them, combine them, move them around, etc. (cf. Z: Scholarly argumentation operates on atomic statements and concepts as fundamental units)
The source of difficulty for chunk decomposition in this case isn't necessarily the fact that the chunks are "tight" in the specific sense that Knoblich meant: that is, the components of the chunk *are* meaningful. It's just that the scholar has to "chisel" them out of the paper. It takes work!
But the point remains that having access to the chunks is important!
Need to be clear about what we mean by Atomicity for Z: Compression is necessary for synthesis: breaking it down to the "concept" level (e.g., a tag, or keyword) is technically a kind of atomicity, but it's... too granular to be useful. The connections explode. I don't think we need to do away with them, but we need higher level blocks, but lower level than a whole document/paper. Right now we only have keywords and papers, and no infrastructure focused on that intermediate layer (I mean, we have standards, but not enough uptake for this to count as infrastructure).