Complete text-only branching SurveyJS scenario ‘The confident answer’, making choices about how to respond to an AI-generated claim and receiving conditional feedback

Interactive scenario

The Confident Answer

Scenario brief

You receive a detailed AI response about new Australian workplace legislation. The AI sounds certain. Your manager needs a decision today. What do you check first?

SurveyJS activity: The confident answer

Node 1: The alert

Your HR colleague forwards an AI-generated summary of "recent Fair Work Act changes affecting redundancy pay." The email claims these changes took effect last week. The summary looks thorough and uses official language.

What do you do?

• Click "Forward to the team immediately. The AI used official terms, so it must be accurate." [Go to Node 2A]

• Click "Check the date and source of the legislative change before acting." [Go to Node 2B]

• Click "Ask the AI to confirm the answer by requesting it again." [Go to Node 2C]

Node 2A: Snap decision

You forwarded the summary. Your manager acted on it during a negotiation. The opposing party's lawyer pointed out the cited provision expired in 2022. The current law differs significantly.

Your colleague's trust in your judgment is damaged. The negotiation position needs rebuilding.

[End: Unsatisfactory. Return to Node 1 or review Lesson 1 Topic 2 on separating claims from evidence.]

Node 2B: Source check

You locate the Fair Work Ombudsman website. The provision the AI cited exists, but the effective date is wrong by eleven months. The dollar thresholds are outdated by two annual indexation rounds.

You now know precisely what requires correction before any workplace action.

[Go to Node 3]

Node 2C: Circular confidence

You prompt the AI again. It generates a similarly confident response with matching sources. Neither run produced a direct link to legislation or the date of royal assent. The AI's consistency masked its shared limitation: no live verification against current registers.

Repeated asking does not substitute for source checking.

[Go to Node 3]

Node 3: The fix

You need to brief your manager in ten minutes. You have identified the claim-anatomy components: the AI stated a legal fact, cited a section number, but supplied no evidence of currency or a checkable source.

What do you include in your verbal briefing?

• Click "The section number and the AI's phrasing, presented as verified." [Go to Node 3A]

• Click "What the AI claimed, what you verified independently, and what remains uncertain." [Go to Node 3B]

Node 3A: Overstated assurance

Your manager repeats your assurance to the board. A director's follow-up question reveals the gap. You must now explain why you treated AI phrasing as verification.

The reasoning gap between claim and evidence was visible but unflagged.

[End: Unsatisfactory. Return to Node 1 or review Topic 4 debrief content.]

Node 3B: Defensible position

Your manager asks follow-up questions. You answer each from evidence you checked, not from AI output. You note two items still requiring legal review before final decision.

Your manager can scrutinise your reasoning. Your decision carries visible support.

[End: Satisfactory. Record completion. Proceed to Topic 4 content debrief.]

SurveyJS configuration

• Completion rule: server-validated upon reaching any End node
• Definition revision: v1.2 (activity variant)
• Accessible fallback: Full text transcript of all node situation text displayed; choice buttons carry aria-label with destination; keyboard-navigable; screen reader announces node transitions
• Branch logic: Node 1 branches to 2A, 2B, 2C; Node 2B and Node 2C converge to Node 3; Node 3 branches to 3A, 3B