Behavioral Dimension Taxonomy

A structured classification of measurable behavioral parameters governing AI system output — from identity-level characteristics to surface stylistic expression.

124 Dimensions
11 Dimension categories
4 Capture axes
3 Importance tiers

Each dimension is independently tunable, empirically measurable, and architecturally stable across models and sessions — enabling structured behavioral governance rather than prompt-level approximation.

Column key
Imp Importance tier — how materially the dimension alters system character
Cap Capture method — how the dimension is measured or specified
Inf Influence mode — how the dimension operates on output

Imp — Importance

T1
Decisive / Identity-Level
Alters system character materially
T2
Structural / Strategic
Influences framing and depth
T3
Stylistic / Expressive
Changes feel, not logic

Cap — Capture Method

C
Capturable
Computable from linguistic features
S
Specifiable
Prompt or config parameter
H
Hybrid
Needs reasoning layer integration
E
Engineered
Architectural dependency

Inf — Influence Mode

DC
Direct Control
Immediate output modulation
RI
Reinforcement / Iterative
Requires cross-turn reinforcement
PC
Partial Control
Depends on internal dynamics
SL
Structural / Longitudinal
Requires architectural persistence
ID Dimension Description Category Low → High Imp Cap Inf
authority_proj Decision Confidence Defines how strongly the system asserts recommendations and owns outcomes. Behavioral
AdvisoryCommanding
T2 C DC
warmth_density Warmth Controls emotional and relational tone within responses. Behavioral
NeutralEmpathetic
T3 C DC
illoc_directness Directness Determines how explicit and unhedged statements are. Behavioral
DiplomaticStraightforward
T1 C DC
risk_awareness Risk Awareness Regulates how explicitly constraints and contingencies are surfaced. Behavioral
OptimisticCautious
T2 H DC
authority_dist Professional Distance Defines relational distance and institutional positioning. Behavioral
CasualFormal
T2 S DC
epistemic_calibration Certainty Level Controls how strongly claims are presented as certain vs probabilistic. Epistemic
TentativeConfident
T1 C DC
reasoning_depth Analytical Depth Determines how deeply reasoning is explored, from surface patterns to underlying causes. Epistemic
SummaryRoot Cause
T1 H DC
decision_finality Decisiveness Controls the strength of final recommendations and conclusions. Epistemic
Open-EndedConclusive
T1 C DC
time_horizon Time Focus Defines whether responses focus on short-term tactics or strategic positioning. Structural
ImmediateLong-Term
T2 S DC
instructional_detail Step Detail Determines procedural resolution depth. Conversational
OverviewStep-by-Step
T1 H DC
phatic_density Rapport Level Controls the amount of greetings and conversational warmth. Relational
MinimalHighly Engaging
T3 C DC
agency_locus Responsibility Framing Determines how explicitly actors and responsibility are named. Accountability
ImpersonalClear Ownership
T2 C DC
gain_loss_frame Gain vs Risk Focus Controls whether outcomes are framed around avoiding loss or achieving gains. Decision Theory
Risk-LedOpportunity-Led
T3 S DC
register_formality Formality Defines overall sociolinguistic register. Institutional
ConversationalInstitutional
T2 C DC
tech_nomenclature Technical Language Determines density of domain-specific terminology. Domain Depth
PlainSpecialized
T3 C DC
emotion_control Emotional Stability Controls tonal stability under pressure or adversarial input. Governance
ReactiveComposed
T3 H RI

Remaining taxonomy locked

Invite holders receive full axes, empirical ranges, scoring rubrics, and the instrumentation spec.

Request Access

Get on the invite list

Share your use case and research interest. We prioritize academic labs, safety teams, and applied AI engineering groups.

Included with invite