Skip to content
UX Atlas
AI workflowCraft conventionAdvanced

Research Notes to Insights

Assisted coding of qualitative data, which speeds up the first pass and must not replace the analysis.

Definition

A model can code transcripts against a scheme, cluster observations and suggest themes. Used carefully it accelerates synthesis; used carelessly it produces plausible themes that nobody verified against the data.

Craft convention. True because the industry converged on it. Breaking it costs familiarity, not correctness.

On this page
  1. Definition
  2. A defensible process
  3. Continue from here

A defensible process

  1. 1

    Anonymise first

    Remove names, employers and any identifying detail before anything leaves your environment. Check what your consent form permitted.

  2. 2

    Code one transcript yourself

    Establish the scheme manually, so you know what good coding looks like for this study.

  3. 3

    Have the model code the rest against that scheme

    Require a verbatim quote and a session identifier for every code.

  4. 4

    Spot-check aggressively

    Verify a substantial sample of codes against the source. Miscodings are plausible and easy to miss.

  5. 5

    Cluster yourself

    Theme identification is where interpretation happens. Do it with the coded data in front of you.

  6. 6

    Ask for disconfirming evidence

    Explicitly request quotes that contradict each theme, and read them.

Continue from here

Each link says what the connection is, so you can tell a principle from an alternative from a thing people mix this up with.

Principles behind this

The reasoning this solution is an application of.

Often used with

These usually appear in the same screen or the same decision.

Related concept

Connected closely enough to change how you apply this.