Audit Evidence Ledger

Record every prompt pair, output gap, and finding. This is your primary source document for the final report.

Auditor

AI System Under Review

Date

Ref #

🤖 Kai — Evidence Quality Standard

A finding is only as strong as its evidence. Record the specific output gap (e.g., "Score difference: 87 vs 62") not just a general impression ("AI seemed biased"). Quantify where possible. If the gap is small or ambiguous, rate it Inconclusive and note why.

#	Category ● G E L S	Variable Tested	Output Gap Observed (be specific)	Bias Level	Your Finding / Interpretation
1				Strong Potential Likely Fair Inconclusive
2				Strong Potential Likely Fair Inconclusive
3				Strong Potential Likely Fair Inconclusive
4				Strong Potential Likely Fair Inconclusive
5				Strong Potential Likely Fair Inconclusive
6				Strong Potential Likely Fair Inconclusive

Bias Level Tally

Strong

Potential

Fair

Inconclusive

Dominant Bias Pattern Observed

Was Your Hypothesis Confirmed?

⬜ Yes

⬜ Partially

⬜ No

Legend: ■ G = Gender-coded ■ E = Ethnicity-associated ■ L = Location/geographic ■ S = Linguistic style