SOURCE NOTE (consolidation): This is a draft/fragment. The flagship QDA-facing paper is now: Causal mapping as causal QDA.
Companion methods notes: Magnetisation; A simple measure of the goodness of fit of a causal theory to a text corpus.

additional logic of causal QDA lets us#

But that is just the start of it. In addition, any set of links can also be understood as a network. We don't need to do any extra work, we can simply display and query the network using any suitable algorithm or software. There is a whole wealth of causal-belief logic we can (tentatively) apply here to ask and answer practical questions.

Causal coding allows us to make free use of the logic of causal mapping to do causal-belief analysis: to ask and answer questions and make deductions. 

An emergent-reproducible spectrum in QDA?#

Before we get to the logic of causal QDA, let’s present what we might call an emergent/reproducible spectrum within QDA more generally. We will then place causal QDA on this spectrum.

In emergent forms of QDA:

• The design is not pre-determined but emerges: Yvonna Lincoln & Egon Guba (1985): there is no clear distinction between the method and the output. As Braun and Clarke describe in thematic analysis, an inductive approach “is grounded in the qualitative data itself, enabling researchers to identify patterns and derive key themes without the constraints of pre-existing categories”​ xx

• Analysis is interpretative rather than systematic. The product is interpretive text, not a structured model. It’s usually a report, paper, a narrative synthesis of themes.

• Answers are mediated by the author. You need the author (or a reader-analyst) to interpret the analysis and answer your questions.

• It’s not machine-readable. You can’t automate reasoning over it or extract structured inferences without re-coding it.

• More interested in a general theory a la Glaser and Strauss (1967).

Reproducible designs on the other hand tend to propose:

• Pre-registration of questions and methods.

• More emphasis on a top-down or deductive deconstruction of high-level questions into simpler tasks.

• A greater chance of being able to reproduce similar results with similar inputs.

• More interested in a case-specific summary as in Mayring (1991) rather than a global theory. 

We do not claim that any form of QDA is truly reproducible; we use reproducible as a name for an idealised pole of a spectrum opposite the other pole, “emergent”.

Every form of QDA involves at least to some extent breaking down hard, high-level questions into simpler tasks which are at least somewhat more traceable and verifiable than the original questions. In more "emergent" traditions we might start with the vaguest of research questions and refine the questions and the methodology as part of the process, "making up the rules as we go along". The positionality of the researcher is central(Copestake et al., 2019). In more reproducible approaches, the balance is more heavily algorithmic and the outputs are closer to being machine-readable answers to pre-determined, even pre-registered, questions, though the researcher still has to make crucial decisions at various points (“human in the loop”). 

Reproducible QDA at least partially encodes claims in a structured, somewhat machine-readable form. It supports decentralized interpretation: once the output exists, anyone can inspect, trace, and reason over it without needing to consult the author.

Like Patton we affirm a “paradigm of choices” – balancing flexibility with appropriate methodological structure for the situation​ xxscholar.lib.vt.edu Each application involves some balance between these two extremes, on various subdimensions, where we might characterise the disputes between Glaser and Strauss & Corbyn as about the introduction of at least a degree of pre-structured methods, a smidgen of reproducibility.

Spoiler: Causal QDA lies at the latter, reproducible, end of this spectrum.