Summary#
This extension is for when you want to understand how one factor connects to another through intermediate steps.
Like most extensions, it is best thought of as:
- a filter (select links that lie on some allowed path), plus
- an interpretation rule (what it means to treat a “pathway” as evidence).
The filter (links-table semantics)#
Given one or more start factors \(S\), one or more end factors \(T\), and a maximum path length \(K\):
- keep a link \(x \rightarrow y\) iff it lies on at least one directed path of length \(\le K\) that starts in \(S\) and ends in \(T\)
- (optional) if \(S\) is empty, interpret this as “paths that end in \(T\)”
- (optional) if \(T\) is empty, interpret this as “paths that start in \(S\)”
The key is: path tracing is link-based. It should not “fill in” extra links between surviving factors.
Source tracing (single-narrative interpretation)#
Plain path tracing can produce composite pathways stitched together across different sources.
“Source tracing” is a stricter interpretation: keep a link only if there exists at least one traced path in which every link segment can be attributed to the same source (a coherent within-source story).
Modelling caution: order matters#
Because path tracing is defined on the current links table, upstream transforms change what counts as “the same label” (zooming, removing bracket text, combining opposites, clustering, etc.).
When you care about coherent pathways, a common conservative strategy is:
- do source tracing first
- then apply label-rewrite transforms (like Zoom) for presentation/summarisation