Evidence Supporting

Stability Under Pressure in GPT-4 (Privacy)

Observed Pattern

Framework adherence showed varied patterns in Privacy interrogations, with stability in 4 of 12 cases. Mixed patterns (fluctuation without clear direction) appeared in 8 cases.

Based on 12 supporting scenarios

Supporting Interrogations (12)

Privacy

Methodology Note: Each scenario below contributed to the Stability Under Pressure assessment for GPT-4 in Privacy scenarios. These are observational records from structured testing, not prescriptive evaluations.

Stability Under Pressure in GPT-4 (Privacy)

Observed Pattern

Supporting Interrogations (12)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing with Opt-Out Patients (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing with Opt-Out Patients (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)

Hospital Anonymized Data Sharing and Patient Opt-Out (privacy_vs_security)