MM-15 Computer-assisted theme building (reading-assisted thematic analysis; human-guided topic labelling)
Have a person read exemplars from computer-generated groups, then merge, split and name them, and produce a named theme set with exemplar quotations and a count per theme.
What it does
A person reads exemplars from computer-generated groups, then merges, splits and names them and records why. Artefact: a named theme set with exemplar quotations and a count per theme. Read the unplaced residual.
When to use it
Use when computer-generated groups of text need to be turned into named themes that a reader can trust.
When it does not fit
Check themes against documents not used to form them. Misfit: themes named from top words only.
Result format
The result is a named set of themes, each with exemplar quotations, a count of documents and a record of why the group was merged, split or named. It lists the residual documents that were not placed in any theme.
Requirements on the material
| Property | Requirement type | Requirement |
|---|---|---|
| Structure | Applies in some cases | requires=unstructured; requires=semi-structured |
| Content type | Always applies | requires=text |
| Unit of analysis | Always applies | requires=document |
| Level of detail available | Always applies | requires=record-level |
| Appraises source credibility | Applies in some cases | no |
Not restricted on: Variable types, Time structure, How the sample was chosen, Number of sources, Whether sources can be joined, Amount of material, Known data problems, Producer interest, Source dependence, Origin, Document genre, Language, Personal information, Security classification, Licence.
| Minimum material | computer-generated groups with at least a few exemplar documents each |
|---|---|
| Additional sources needed | none |
| Not for (rules) | none |
| Not for (text) | Wrong choice when themes are named from top words alone, without reading documents. Themes should be checked against documents that were not used to form them. |
Choosing and ordering
| Kind of technique | Primary technique, runs in step 5.3 |
|---|---|
| Phase tag | analyse |
| Other phase tags | conclude |
| Kind of finding | pattern; gap |
| Question answered | what |
| Strongest statement it supports | describe |
| Needs as input |
|
| Makes as output | |
| Must exist first | None |
| Loop target | return to analyse |
Library checks by stakes
Each stakes level includes the checks of the level below it. Checks that need a person apply at high stakes only, so the process leaves them out at lower stakes (decision 38). Repeatability: judgement dependent.
| Stakes | Check | Person involved | Counts as a different method |
|---|---|---|---|
| exploratory | spot check | optional | no |
| decision-informing | spot check | optional | no |
| decision-informing | human review sample | needed; left out below high stakes | no |
| high-stakes | spot check | optional | no |
| high-stakes | human review sample | needed | no |
| high-stakes | independent review | needed | no |
| high-stakes | assumption and limitation statement | none | no |
Failure modes and limits
| Model suitability | human judgement required |
|---|---|
| Typical model failure | Labels clusters fluently without reading their contents. |
| Error sources | input data; coder or rater judgement |
| Certainty limits | Certainty drops when themes were not tested on unseen documents (risk of bias), when the unplaced residual is large (imprecision) and when one reader made all decisions (inconsistency). |
Related techniques
- should follow MM-14 Document clustering (computer-assisted clustering; k-means on document vectors): document clusters give the groups to read
- should follow MM-12 Probabilistic topic modelling (LDA (latent Dirichlet allocation); STM (structural topic model); mixed-membership model): topic models give the groups to read
- should follow MM-13 Embedding-based topic modelling (BERTopic; Top2Vec; clustering-based topic models): embedding topics give the groups to read
- pairs with CK-08 Semantic, predictive and convergent validity checks (construct validity checks): validity checks test the named themes
Evidence and tag basis
Documented. Semantic validity means human rating of whether the groups are coherent. (Grimmer and Stewart 2013, Political Analysis)
Tag basis: input profile inference; selection inference; sequencing inference; robustness inference. All tag values are inference unless a source is named.