Purpose
Use this lesson to code open-text responses consistently and retain evidence for each theme. The aim is to produce work that a colleague can review and act on, not merely text that sounds convincing.
Lesson Explanation
Scenario: Customers mention slow setup, pricing, and helpful support.
Better instruction: Group responses into themes, count each theme, provide two representative quotes per theme, and label any response that fits multiple themes.
Thematic analysis needs a coding rule, not a cloud of keywords. Ask for theme definitions, counts, supporting excerpts, and treatment of mixed responses. Review the labels on a sample before using them for decisions. Themes have counts and quotes, and mixed responses are identified.
Work in a short cycle: provide the relevant input, state the result and limits, inspect the draft against the acceptance criteria, then revise the instruction when a requirement is missing. Keep the final human decision with the person responsible for the work.
Practice Questions
-
A prompt asks an AI to “find the themes” in survey responses with no further guidance. What does this lesson say thematic analysis needs, instead of just a general request?
View Answer
A coding rule, not a cloud of keywords – theme definitions, counts, supporting excerpts, and treatment of mixed responses.
-
What does this lesson recommend doing with a response that fits more than one theme?
View Answer
Labeling it as fitting multiple themes, rather than forcing it into a single category.
-
Write a prompt for analyzing survey feedback about a mobile app, following this lesson’s guidance.
View Answer
Something like: “Group responses into themes, count each theme, provide two representative quotes per theme, and label any response that fits multiple themes.”
-
A thematic analysis produces theme labels but no supporting quotes or counts. What does this lesson say is missing?
View Answer
Counts and representative excerpts for each theme, which support and validate the identified themes.
-
What does this lesson recommend doing with the AI’s theme labels before relying on them for a business decision?
View Answer
Reviewing the labels on a sample before using them for decisions.
-
A survey response mentions both a pricing complaint and a product feature request, but the AI forces it into just the “pricing” category. What does this lesson say about this outcome?
View Answer
It should have been labeled as fitting multiple themes, rather than forced into a single one.
-
A completed thematic analysis includes theme names, counts for each, two representative quotes per theme, and flags for mixed responses. What does this reflect?
View Answer
A complete thematic analysis following this lesson’s full recommended structure.
-
Why does this lesson describe raw thematic output without structure as “a cloud of keywords”?
View Answer
Without theme definitions, counts, and supporting evidence, loosely identified themes lack the structure needed to be reliably interpreted or acted upon.
-
A prompt for survey analysis requests theme counts but doesn’t ask for any supporting quotes. What is missing, per this lesson’s guidance?
View Answer
Representative excerpts (quotes) for each theme.
-
Why might skipping the step of “reviewing labels on a sample” before broader use be risky, according to this lesson?
View Answer
Unreviewed labels could reflect a misinterpretation of the data, and using them broadly without a sanity check risks basing decisions on flawed categorization.
-
A team reviews a sample of ten labeled responses and finds the AI consistently mislabeled negative comments as neutral. What earlier step from this lesson helped catch this issue?
View Answer
Reviewing the labels on a sample before using them for decisions.
-
A thematic analysis of customer complaints identifies five themes, each with a count and two quotes, with three responses flagged as fitting multiple themes. What does this reflect?
View Answer
A well-structured thematic analysis consistent with all the elements this lesson recommends.