Design a contextual observation protocol
Prepare a neutral field-observation sheet that captures behaviour, environment, workarounds, and service context consistently.
Why
The workflow problem
Teams know whom to interview but fail to capture behaviour, environment, workarounds, and service context consistently.
AI can turn approved research questions into a first-pass observation template and identify coverage gaps. Its suggestions are not observations; researchers must secure consent, notice context, record what happened, and separate description from interpretation.
Concrete output
What you will produce
Field-observation sheet with behavioural prompts, context checklist, note template, consent plan, and data-handling rules.
How
Run the workflow
Set the observation boundary
The researcher defines the objective, setting, activities, observation window, consent approach, data minimisation rules, and prohibited capture. Check that observation is necessary and proportionate.
Draft neutral observation prompts
Using only the approved boundary and research questions, AI drafts prompts for observable behaviour, environment, tools, interruptions, workarounds, and exact quotes, plus a separate interpretation field.
Observation-sheet draft
Convert the approved research questions into a field-observation sheet. Use neutral, observable prompts; separate “what happened” from “possible interpretation”; include context, tools, interruptions, and workarounds. Do not suggest covert observation, collect personal data not named in the brief, or label motivations as facts.Review consent and language
The researcher and privacy owner remove unnecessary fields, check that prompts do not invite inference about protected or sensitive attributes, and confirm a clear participant explanation and withdrawal route.
Rehearse in the setting
The researcher tests the sheet in a low-risk rehearsal, checks note-taking feasibility and visibility, then issues the final version. During fieldwork, observers record exact behaviour and quotes before forming interpretations.
Human–AI partnership
Who contributes what
AI contribution
Produces a provisional template and coverage check from an approved brief; it neither observes nor validates behavioural claims.
Human responsibility
The researcher chooses the method and setting, secures consent, manages identifiable data, observes context, and distinguishes observed fact from interpretation.
Stop rule
Stop if AI suggests covert observation, capture of sensitive data without a lawful and consented basis, or interpretive labels presented as observations.
Check before use
Watch out
A comprehensive-looking template can crowd out emergent detail. Leave room for exact notes and surprises, avoid recording bystanders unnecessarily, and never replace field observation with an AI-generated account.
Evidence base
What supports this workflow
Contextual observation is a well-established human-centred design method. The proposed AI role is a practical drafting hypothesis rather than a directly evaluated workflow in the harvest, so it should be piloted and assessed locally.
- Stanford d.school · Ethnography Expedition
Supplies a canonical field-research sequence built around observation, learning, insight development, and reframing.
- Government Digital Service · User research in discovery
Names observation of current tasks and barriers as a discovery activity and situates it in end-to-end service context.
- Government Digital Service · Plan a round of user research
Informs planning, access needs, consent, recording, and fieldwork logistics that the protocol must address.