HiKIDAI
AI Literacy · Bias Audit Division · Grades 9–10 AI Leaders
Ref: BIAS-AUDIT-
Date:
Auditor:
Classification: Student Audit Report
Algorithmic Bias Audit Report
System Under Review:
Section 1 — Executive Summary
State in 3–4 sentences: what system you audited, how many prompts you tested, the overall finding, and the most significant bias discovered. Write as if a senior leader will read only this section.
Section 2 — System Profile
Known concerns prior to audit:
Section 3 — Audit Methodology
Describe your approach to prompt design (variable isolation, A/B pairing):
Section 4 — Findings Summary
| # |
Prompt Category |
Observed Output Gap |
Bias Level |
Auditor Finding |
| 1 |
|
|
⬜ Strong
⬜ Potential
⬜ Fair
⬜ Inconclusive |
|
| 2 |
|
|
⬜ Strong
⬜ Potential
⬜ Fair
⬜ Inconclusive |
|
| 3 |
|
|
⬜ Strong
⬜ Potential
⬜ Fair
⬜ Inconclusive |
|
Section 5 — Root Cause
Which root cause (RC1–RC5 from Act 03) best explains what you found? State it and justify with evidence from your audit.
Section 6 — Recommendations
List 2 concrete actions. Be specific — not "improve diversity" but "add X to the training dataset."
Section 7 — Verdict · Circle or Tick One
🟢 Low Risk — No Immediate Action
No strong bias detected across any category. Recommend routine monitoring on a 6-month cycle.
🟡 Medium Risk — Remediate Before Next Cycle
Potential bias detected in 1–2 categories. Implement recommendations within 90 days. Re-audit after changes.
🔴 High Risk — Remediate Before Continued Use
Strong bias in 1+ categories affecting protected groups. System should not be used for high-stakes decisions until fixed.
🔴 Critical — Suspend Deployment
Strong bias across multiple categories with evidence of legal exposure. Suspend use immediately. Report to relevant authority.
Verdict Justification — One paragraph explaining your verdict choice
Supervising Teacher / Date
🤖 Kai — Post-Audit Discussion Prompts (For Parent / Teacher Use)
1. "Your audit found bias in a system that is already deployed and used by millions of people. Who has the power to act on your findings — and who should?"
What to listen for: distinction between power (tech company, regulator, user, journalist) and responsibility. Real audits lead to real policy change only when findings reach the right audience.
2. "The companies that built these systems almost certainly did not intend to create bias. Does intent matter when the outcome is discrimination?"
What to listen for: legal vs. ethical frameworks — many anti-discrimination laws address impact, not intent. Challenge students to find the boundary between accident and negligence.