Example 1 · An earlier AI-and-projects coding draft
A theme is a claim about the material
The earlier prototype arranged ideas about AI and project work into four themes. Explore its properties, subcategories and proposed ranges. Select a label to inspect the annotation that was offered in support of it.
Preserved for inspection, not authenticated as research. All 73 labels and 62 annotations come from the old public prototype. Its claims of 919 analysed posts, a 35-fold increase and manual coding of every post had no accompanying corpus, coding log or calculation. The annotations were called LinkedIn quotations, but post links were not supplied. They are retained here as unverified draft text, not attributed quotations or established findings. Dates and product names inside them belong to that earlier draft.
Try a critical reading
Open Practice Evolution Trajectory → Properties → Change Velocity. Does a remark about reconstructing images from MRI scans support a claim about accelerating adoption in project management? The old draft links them, but the link needs an argument. Likewise, a tool launch is not evidence of organisational adoption.
The four themes can be useful questions: How does work change? What happens to professional identity? How is learning shared? How far does a change spread? The draft does not establish a universal progression from traditional to AI-native practice or show that orchestration produces a competitive advantage.
Example 2 · A selected Legends Tower text sample
Positive words do not necessarily mean support
Explore the old prototype’s 22 selected news and social-media records about the proposed Oklahoma City tower. Compare its supplied stance label, interpreted intention and sentiment scores. This is a historical method example, not a current project update or a representative opinion survey.
Keep the units separate. These are statements, not 22 independent stakeholders. Some people and threads recur. “Supporter”, “Opponent” and “Neutral” are the original analyst’s labels for the selected items, not verified permanent roles. A safety assessment or doubt about feasibility does not establish opposition to the whole project.
Counts by original stance label
Changing a filter changes these counts. They describe only the selected records.
Recorded score averages by label
The TextBlob-labelled polarity (−1 to +1) and integer lexicon scores were supplied by the old file. Its scoring version, lexicon and execution trace were not included. They are not recomputed or validated here; the two scales are not comparable magnitudes.
Recorded polarity and lexicon score
Each point is a statement. The vertical scale is the old lexicon score; the horizontal scale is recorded polarity. Select a point or a table row to inspect the text. Jitter separates overlapping points without changing displayed values.
Statement → interpreted intention
Each pair retains one source record and its earlier interpretation. The arrow says the draft interpreted this statement as…; it is not an ontology-backed assertion about what a person intends. Repeated names remain separate statements.
Inspect a statement
Select a point, interpretation pair or record ID.
Try: a mismatch that matters
Search purpose. The questioning post has a positive recorded polarity but an “Opponent” label. Search hazard: the safety statement was labelled “Opponent” while polarity was zero. Neither numerical tone nor an analyst’s label alone settles stance, intention or factual correctness.
Two news records—news_03 and news_04—have text and attribution checked against their cited 4 June 2024 article. The other attributions remain unverified in this review. This limited check does not validate the scoring, sampling or original intention labels. Missing social-post dates remain unknown.
From text to a defensible interpretation
Qualitative coding
A useful coding structure makes its examples and proposed relationships inspectable. Grounded-theory research asks more of the process than arranging labels into a tree. Diehl and colleagues’ study of visualization guidelines, for example, describes explicit coding, further sampling, negative-case analysis and additional empirical studies. This draft provides no equivalent trail.
When reviewing a theme, ask what text supports it, what context is missing, what contradicts it and whether a simpler interpretation fits. “Axial coding” concerns examining relations among categories; a parent–child drawing alone does not establish those relations.
Tone, stance and intention
TextBlob’s documentation distinguishes polarity and subjectivity scores. Neither is a truth score or a direct measurement of support for a specific proposal. Here, the scoring pipeline is unavailable; keep these numbers as historical outputs to question.
A statement can favour housing, question financing and raise a safety condition at once. Identify the object and time of each claim before grouping people as supporters or opponents. Selection, duplicated sources and missing dates matter before averages can support a conclusion.
Questions retained from the earlier recommendations
The original qualitative draft proposed small pilots, skills development, peer learning, explicit human review and reflection on changing roles. These remain sensible topics to investigate, not conclusions demonstrated by this sample. Its fixed 30-day/6-month programme and competitive-necessity claims have been removed.
Keep the inspected material
Export the complete coding tree and the currently filtered statement records, with their evidence limits. This page does not store changes or upload data. The original examples had no saved user state to migrate.