Contextual Privacy Management Framework

Mari Privacy Profile Index

Adaptive privacy profiling for stakeholder demos

CPMF

Five signals for contextual privacy management

CPMF frames disclosure behavior as a five-axis profile. Each axis changes how an agent should decide what to reveal, to whom, and under what conditions.

Control Orientation

Approval workflow

Measures how strongly the user wants to retain direct approval rights over disclosure decisions instead of delegating them to the AI agent.

LowComfortable delegating routine disclosure decisions to the system.
HighWants explicit review and approval before disclosure.

Openness

Disclosure appetite

Combines baseline disclosure willingness with impression-management sensitivity, capturing how willing the user is to share even when reputational risk is present.

LowDiscloses selectively and manages reputation carefully.
HighShares broadly, including potentially vulnerable or unflattering details.

Domain Selectivity

Boundary setting

Captures how strongly the user wants to protect non-professional information such as hobbies, feelings, values, relationships, and private routines.

LowPresents a whole-person identity with weak compartmentalization.
HighMaintains strong professional boundaries around personal-domain information.

Self-Exploration Orientation

Inference visibility

Tracks whether the user wants the agent to uncover patterns, traits, or signals they have not yet consciously recognized and turn them into a reflective dialogue.

LowPrefers speculative inferences to stay in the background unless requested.
HighActively wants the system to surface hidden patterns and inferred insights.

Contextual Sensitivity

Rule granularity

Reflects how strongly privacy preferences depend on context, aligning with contextual integrity, purpose limitation, and temporal persistence concerns.

LowPrefers stable global rules that apply across most situations.
HighUses different disclosure rules across contexts, purposes, and retention windows.