SafeBARS

Encounter stress-testing workspace

Research rehearsal
1
Collect information
Guided intake available
2
Map the encounter
Awaiting materials
3
Trace breakdowns
Not run
4
Decide and hand off
No issues
Encounter Orchestrator
Waiting for protocol materials.
Breakdown Scenario Agent
No scenario queue.
Relationship and Power Agent
No relationship check.
Boundary and Handoff Agent
No boundary review.
Inspectable audit plan Material-routed tasks, tools, dependencies, and stop conditions
Build an encounter map to let the Orchestrator propose a bounded task plan.
No encounter map yet
Add protocol materials or load the sample, then build the map.
No scenario results yet
Select difficult situations in the encounter map, then choose “Run selected traces”.
No researcher decisions yet
After a scenario test finds a protocol gap or ambiguity, it will appear here with a proposed change.
Expert review is not needed yet
Run the audit first. If a question requires human judgment, SafeBARS will create a review request here.

Quick guided protocol intake

Step 1

How to use SafeBARS

Researcher workflow, expert review, and record continuity
SafeBARS is a staged workflow. Expert input does not appear immediately after intake. First create the protocol case and run the audit; expert fields appear only when the AI creates a human handoff.

For researchers

Prepare and stress-test a sensitive human-facing protocol.

1
Complete Quick Intake

Answer six core questions. AI projects receive one additional governance question.

2
Review the populated fields

Correct the protocol materials in the left panel before analysis.

3
Build encounter map

Create the case, index passages, and inspect the participant journey.

4
Run selected traces

Choose breakdown scenarios and run the audit to identify issues and human handoffs.

5
Open expert review

When handoffs exist, the header changes to “Open expert review (N)”. Use it to preview the expert form, or copy the invitation for a real expert.

6
Respond and export

Link protocol revisions to expert requests, then export the ethics application, research design, or audit report.

For ethics experts

Review only the questions the AI cannot responsibly settle.

1
Open the invitation link

A valid case-specific invitation is required. The generic caseload is not an input form.

2
Identify your role

Enter your reviewer name and choose the most appropriate expert role.

3
Enter expert response

Write the required change or advice, ethical rationale, and closure condition.

4
Choose a workflow action

Save advice, request clarification, redirect, or mark the handoff addressed.

5
Review revisions

Use History to compare researcher responses with earlier expert requests.

Plain-language guide to interface terms

Source passageA numbered excerpt from material you submitted. For example, “Interview questions · passage 3” comes from internal ID INT-003.
Encounter mapThe participant journey SafeBARS reconstructs from recruitment through follow-up.
TraceA step-by-step rehearsal of one possible difficult situation.
IssueA protocol gap or ambiguity for the researcher to accept, edit, reject, or defer.
Handoff / expert review requestA question the AI cannot responsibly settle and sends to a qualified person.
Documented / partial / missingWhether the submitted materials contain enough evidence for a framework dimension. These are coverage labels, not ethics scores.
Keep your record: use the in-page “Open expert review” button and “Researcher workspace” return link in the same browser tab. The current prototype stores access credentials in that tab. Closing the tab, clearing browser storage, or opening the generic expert caseload directly may remove the local route back to the case. Export important records before ending a session.
Working
SafeBARS is updating the shared audit state.