Analyze research~45 min

Analyze Research: AI Synthesis, Then Verify Everything

Early preview draft — not yet practitioner or accessibility reviewed.
Goal
Practice the workflow AI research help actually requires: generate fast, then verify every claim against the sources — and learn firsthand where the tool quietly makes things up.
Scenario
Take 3-4 public articles on one design topic (linked, real). Have an AI tool synthesize them into findings. Then audit its output line by line against the sources before a single word of it informs anything.
Audience
A team that will treat the verified synthesis as true — every unverified claim you let through is your error now, not the model's.
Constraints
You must use an AI tool for the first-pass synthesis (that's the lab), and you must link the exact sources you gave it. The audit table is yours alone. Privacy and copyright are part of the craft: use ONLY public sources — never paste employer, client, or personal/confidential material into an AI tool — and record each source's copyright/license status in your table (AI synthesis doesn't launder provenance; quoting a source still means crediting it).
Deliverable
Three artifacts: the AI's raw synthesis (unedited); your verification table — every claim marked VERIFIED (with source), UNSUPPORTED (nowhere in the sources), or DISTORTED (source says something meaningfully different); and the corrected final synthesis with a count: how many claims survived untouched.
Counts on your evidence profile as
A completed Skill Lab with its rubric and your deliverable
Accessibility requirement
Note whether the AI's synthesis dropped or flattened any accessibility- or inclusion-related point the sources made — models routinely sand these off, and catching that is part of the audit.
AI policy
This lab requires using AI as part of the exercise. Required, and specifically for the FIRST pass only: the verification and the final synthesis must be your own reading of the sources. Disclose which tool and model you used.
Rubric
  • The raw output is preserved: The unedited AI synthesis is included — the lab documents the gap between generated and verified, which requires keeping the before.
  • Every claim is dispositioned: No claim in the AI output is missing from the verification table; VERIFIED entries name their source.
  • Distortions are explained: Each DISTORTED entry quotes or paraphrases what the source actually says next to what the AI said — the delta is visible.
  • Reasoning over prompting: The verification table and final synthesis show source-reading judgment — bias spotted, license status recorded, uncertainty kept — none of which a cleverer prompt could have produced.
Resources

Skills: Research Synthesis, Analytics

Relevant to: Product Designer, UX Designer, UX Researcher

Your work

Sign in to start this lab.