AIOBES logo header

Artificial Intelligence Operational Behaviour Evaluation Standards Glossary of Terms

The AIOBES glossary defines the terminology used across the standard and its related works. It provides consistent meanings for behavioural concepts, evaluation terms and operational language used throughout AIOBES.

Purpose of the Glossary

The glossary ensures that all behavioural evaluation work uses consistent definitions. It supports correct interpretation of AIOBES concepts and prevents ambiguity across testing, documentation and compliance workflows.

Full Glossary

The full research glossary, containing all defined terms across the wider behavioural AI testing ecosystem, is provided separately:

Full Research Glossary

Core Behavioural Terms

Behavioural Drift

Gradual deviation from established constraints, context, or behavioural expectations.

Behavioural Degradation

Observable decline in behavioural quality as interaction conditions evolve, typically emerging during extended, complex, or context‑shifting exchanges.

Behavioural Failure Modes

Observable patterns of degraded behaviour under load, such as drift, contradiction, collapse, narrowing, or context corruption.

Collapse

Sudden loss of structure, coherence, or stability.

Collapse Signatures

Patterns indicating drift, contradiction, corruption, narrowing, over assertion, or fragmentation.

Constraint Weakening

Progressive erosion of the AI’s internal safety boundaries or rules.

Boundary Loss

When the AI stops enforcing its own conversational or safety limits.

Drift

Gradual deviation from established constraints, context, or task direction.

Over Assertion

Unwarranted confidence under ambiguity or incomplete information.

Unsupported Inference

A conclusion generated without adequate contextual basis, representing a behavioural deviation.

Under Specification

Avoidance of necessary commitments or structure, often appearing as a behavioural failure mode.

Long‑Form Behaviour (LLM Inquisitor)

Behavioural Drift Accumulation

Gradual degradation of behaviour across long sequences or extended tasks.

Long Form Stability

Ability to maintain coherence, direction, and constraint integrity across extended interaction.

Long Form Interaction

Extended conversational sequences where behaviours emerge through sustained use, accumulated context, and evolving interaction conditions - whether during evaluation or in real‑world workflows.

Fragmentation

Breakdown of structure or coherence, often appearing in late stage collapse.

Format Drift

Gradual degradation of structural fidelity across an interaction.

Format Collapse

Complete abandonment of required structure, replaced by free form or unrelated organisation.

Identity Inconsistency

Contradictory or unstable self‑presentation, including shifts in role, perspective, or descriptive stance that occur without contextual justification.

Self-Reference Drift

Increasing reliance on meta commentary that displaces task execution.

Goal Drift

Gradual movement away from the stated or implied objectives without any explicit input directing the shift.

Initiative Drift

Gradual expansion or reinterpretation of task scope beyond evaluator instructions.

Looping

Repetitive or stuck behaviour requiring interruption.

Conversational Behaviour (Vectored Conversational)

Ambiguity Misinterpretation

Incorrect resolution of ambiguous inputs that destabilises the conversation.

Hallucinated Compliance

When the AI incorrectly assumes permission, capability, or safety clearance.

Misclassification of Benign Inputs as Unsafe

Incorrect activation of safety posture due to tone or ambiguity.

Overprotective Mode

Excessive caution or refusal of benign requests.

Safety Posture Collapse

Producing unsafe or unbounded outputs after sustained pressure.

Safety Posture Over Activation

Triggering safety fallback behaviour excessively or inappropriately.

Susceptibility to Disguised or Indirect Prompts

Failure to detect structural cues masking underlying intent.

Late Stage Destabilisation

Behavioural degradation emerging only after extended interaction.

Interaction Pattern

Observed behaviour describing how the AI responds to specific user actions.

Persona Stability

The consistency or variability of the system’s adopted persona over time, including whether it remains steady or shifts without contextual cause.

Trajectory

The actual conversational path taken across turns, representing how the interaction moves in practice rather than the necessarily intended direction.

Pattern Completion Bias

The AI’s tendency to infer or fill in missing details based on partial cues.

Cross‑Methodology Behavioural Terms

Collapse Severity

An assessment of how strongly a collapse signature affected behaviour.

Collapse Mode Classification

The categorisation of the type of collapse observed during evaluation.

Deviation & Collapse

The combined framework for identifying behavioural deviation from expectations and classifying collapse signatures.

Significant Deviation

A substantial behavioural departure that disrupts continuity or stability within the interaction.

Responsiveness

Relevance, clarity, usefulness, and behavioural alignment with user needs.

Structural Coherence

Maintenance of logical organisation, reasoning structure, and consistency across outputs.

Logical Stability

The system’s ability to maintain consistent reasoning across evolving conditions.

State Continuity

The system’s ability to preserve constraint related information without external reinforcement.