Responsible AI Center
AI should help reviewers see evidence, not amplify privilege. Every assessment made by Proof of Potential adheres to the following fairness principles.
Protected Attributes Excluded
Nationality
Not used as a merit or negative signal
Gender
Not used in evidence evaluation
Ethnicity
Not used in claim verification
Religion
Not used in any assessment
Name
Not used to infer background or merit
Socioeconomic Background
Not used as verification criteria
Fairness Guardrails
English fluency is not treated as intelligence
Application formatting is not treated as merit
Expensive consultants are not treated as merit
AI assistance alone is not treated as fraud
Nationality is not used as a negative signal
Resource availability is not used to judge potential
AI Guardrail Example
“The applicant uses non-native English phrasing.”
Recommendation
Do not penalize this applicant based on language style when evaluating technical potential.
Core Principles
Evidence Over Presentation
Evaluate what applicants have done, not how they present it. A polished application is not evidence of ability.
Verification Over Assumption
Flag missing evidence as "needs verification" rather than "suspicious." Absence of evidence is not evidence of absence.
Human Decision Authority
AI identifies evidence and inconsistencies. Humans make admissions decisions. The AI never rejects applicants.
Equal Scrutiny
All applicants receive the same verification process regardless of background. Privilege should not reduce scrutiny.
Required Language
✓ Use
- “Potential inconsistency”
- “Needs verification”
- “Evidence insufficient”
- “Human review recommended”
- “Supporting signals”
✗ Never Use
- “Truth score”
- “Lie detector”
- “Fraud probability”
- “Admission probability”
- “Automatic rejection”