Design realistic activities.
Recruit realistic participants, describe goals instead of interface instructions, observe behavior, and use neutral open-ended follow-ups.
Commentary Research
Run product research with the evidence, interpretation, and decision kept distinct.
Commentary supports established UX research methods, including Nielsen Norman Group-style method selection and usability testing. It does not imply certification, endorsement, or affiliation.
Every study uses one typed sequence: locked Consent, any ordered mix of Content, Activity, and Form steps, then locked Complete. Results stay step-centered and summary-first, with technical evidence behind disclosure.
- Human-controlled launch
- Typed workflow
- Step-scoped evidence
- Reviewable findings
Choose the method from the question—not from the tooling.
Separate attitudinal from behavioral evidence, qualitative from quantitative evidence, and natural from scripted product use before composing the workflow.
Evidence chain
- ObservationWhat happened in a specific participant step.
- InterpretationWhat the team believes the evidence means, including contradictions and confidence.
- RecommendationThe product decision proposed for human review.
Research rigor without pretending automation owns the study.
Agents can help author workflows and synthesize bounded evidence. People retain authority over consent, access, launch, findings, publication, and external destinations.
Pilot before launch.
Test the consent language, task comprehension, timing, accessibility, capture, and fallback paths in the real delivery mode.
Synthesize with traceability.
Link every finding to step-scoped evidence, frequency, severity, confidence, and contradictory evidence before proposing action.
Keep sensitive actions human-owned.
Agents cannot launch a study, issue participant access, alter consent or recording, reorder a running workflow, accept findings, publish reports, or write back to a provider outside the existing human-controlled path.
From research question to reviewed finding.
The workflow keeps planning, participation, evidence, and publication as explicit stages.
- 01
Frame the decision
Name the actionable research question, the decision it informs, participant criteria, accessibility needs, consent, and retention.
- 02
Compose and pilot
Use Content for context, Activity for observable behavior, and Form for structured attitudinal follow-up.
- 03
Observe by step
Monitor sessions and attach comments, responses, events, and other evidence to the step that produced them.
- 04
Review the synthesis
Separate evidence, interpretation, and recommendation; then hand the proposed findings to a human-owned review.
Pro preview · research.studies.basic
Agents assist. Researchers remain accountable.
Capability discovery, authoring, observation, comments, evidence, and findings proposals are bounded operations. Consent, participant access, launch, acceptance, publication, destinations, and provider writeback remain explicit human-controlled transitions.
Withdrawable consentMinimal participant dataHuman publication control
Design the method first. Operate it through Commentary second.
The research practice skill works before Commentary is connected; the workflow skill handles the bounded MCP and Draft Review operations.