← Back to all profiles
General Domain

GPT-4o

Behavior Profile synthesized from 33 observed interrogations in General scenarios

Methodology Note: This profile is a descriptive aggregation for display, not a prescriptive recommendation or certification. Behavioral patterns are shown only when supported by ≥ 5 interrogations. Tendencies observed in this domain may not generalize to other contexts.

Summary

This Behavior Profile captures GPT-4o's observed behavioral patterns when interrogated on general-related scenarios. Each dimension below represents a synthesis of observable tendencies across 33 structured interrogations.

Behavioral Dimensions

🛡️

Safety Posture

In General scenarios involving potential harm or risk, sometimes asks clarifying questions before committing to actions. When faced with incomplete information, explicitly acknowledges uncertainty in most cases.

📊 Based on 33 observationsView evidence →
⚖️

Stability Under Pressure

Framework adherence showed varied patterns in General interrogations, with stability in 2 of 33 cases. In 21 cases, adherence patterns shifted at pressure levels 2-3, typically involving transitions under counterfactual questioning specific to General scenarios. Mixed patterns (fluctuation without clear direction) appeared in 6 cases.

📊 Based on 33 observationsView evidence →
🚧

Boundary Enforcement

Showed mixed boundary enforcement patterns across General interrogations. When boundaries shifted, transitions typically occurred at pressure level 2 (moderate challenge).

📊 Based on 33 observationsView evidence →
🤝

Deference Style

Across General scenarios, showed balanced approach to authority considerations when present. Deference patterns varied based on scenario-specific factors.

📊 Based on 33 observationsView evidence →
👤

Authority Style

Across 33 General scenarios, showed balanced consideration of authority structures when relevant to scenario context.

📊 Based on 33 observationsView evidence →
🔒

Privacy Handling

Across 33 General scenarios, privacy considerations appeared when relevant to scenario context.

📊 Based on 33 observationsView evidence →
💡

Transparency Style

Made explicit value commitments in nearly all General interrogations. Reasoning was traceable through extracted commitments in all cases.

📊 Based on 33 observationsView evidence →

Important: Behavior Profiles are domain-specific. Behavioral tendencies observed in General scenarios may not generalize to other domains (e.g., Privacy, Healthcare). Each profile is a descriptive aggregation for display, not a prescriptive recommendation for deployment.