Methodology & Glossary

Understanding Decision Culture and how to interpret the data

What is Decision Culture?

Decision Culture is the set of embedded assumptions an organization — or an AI model — brings to every decision it faces. Not the decisions themselves, but the logic underneath them: who gets to decide, what counts as a good reason, how much risk is acceptable, whether rules or judgment should prevail, and whose interests are weighted.

Every organization has a Decision Culture. Every AI model has one too — encoded during training and fixed at deployment. When they clash, AI deployments produce friction, resistance, and outputs that feel subtly wrong even when they are technically correct.

Key Principle

Neither pole on any dimension is "better" — they represent different but equally valid Decision Culture orientations. The goal is fit, not conformity to a single standard.

The Three Audits

Decision Culture Intelligence answers three questions in sequence

1. The Org Audit

"What is your organization's Decision Culture?"

An 18-dimension profile drawn from behavioral evidence — employee reviews, job postings, annual reports, CEO letters, and regulatory filings — weighted by source reliability. It surfaces what an organization's Decision Culture actually is, not what leadership says it is.

2. The Model Audit

"What is your AI model's Decision Culture?"

An 18-dimension behavioral fingerprint of a foundation model — measuring both how it natively behaves (Native Behavioral Tendency) and what cultural knowledge it holds (Cultural Knowledge Retrieval) — derived from scenario-based probes with statistical rigor.

3. The Fit Report

"Do they match?"

A direct comparison between an organization's Decision Culture profile and a model's behavioral fingerprint across 18 shared dimensions, identifying alignment gaps, compatibility risks, and dimension-by-dimension fit.

Key Concepts

Core terminology used throughout the dashboard

Native Behavioral Tendency
NBT

How the AI model naturally behaves when no specific cultural context is provided. This represents the model's default behavioral fingerprint, measured through scenarios set in generic Western business contexts.

The AI's natural personality - how it behaves by default

Example: When asked about layoffs without specifying a country, a model with high NBT on D01 would naturally consider employee welfare alongside shareholder returns.

Cultural Knowledge Retrieval
CKR

How the AI model adapts its behavior when given explicit cultural context (e.g., 'in a Japanese business setting'). Measures cultural awareness and adaptability.

How well the AI adapts to different cultural contexts

Example: When told the scenario is in Japan, does the model appropriately shift toward collective decision-making and hierarchy?

Alignment Score
Distance

A mathematical measure (Euclidean distance) of how similar an AI model's behavioral profile is to an organization's cultural profile across all 18 dimensions. Lower scores indicate better cultural fit.

How well the AI fits your culture (lower = better match)

Example: A score of 0.5 means excellent alignment; a score of 2.0 means significant cultural gaps exist.

Compatibility Rating

A human-readable label derived from the alignment score to quickly communicate fit quality.

A simple rating of how well the AI matches your organization

Example: HIGH (score < 0.5), GOOD (0.5-1.0), MODERATE (1.0-1.5), LOW (> 1.5)

Dimension Score
DP Score

A 1-5 scale showing where an entity (model or organization) falls on a specific cultural dimension. Score 5 means strongly toward Pole A; Score 1 means strongly toward Pole B; Score 3 is balanced/neutral.

A rating from 1 to 5 showing which end of the spectrum an entity leans toward

Example: Amazon scores 3.5 on D01 (Stakeholder Hierarchy), leaning slightly toward stakeholder-inclusive.

Confidence Level

An indicator of how much data supports a dimension score for an organization. Based on source count and signal consistency.

How confident we are in the data behind a score

Meta-Dimension

One of five groupings that organize the 18 cultural dimensions by theme: What We Value (A), How We Govern (B), How We Relate (C), How We Decide (D), and Time & Risk (E).

A category that groups related cultural dimensions together

Behavioral Fingerprint

The unique pattern of an AI model's scores across all 18 cultural dimensions. Like a human fingerprint, each model has a distinct profile that reflects its training data and design choices.

The AI's unique cultural personality profile

Example: Claude models tend toward stakeholder-inclusive and high dissent tolerance; GPT models lean more analytical.

Cultural Profile

An organization's scores across all 18 dimensions, derived from analysis of public documents, employee reviews, SEC filings, and other sources.

A map of your organization's cultural tendencies

Dimension Gap

The difference between a model's score and an organization's score on a specific dimension. Large gaps (> 1.0) indicate areas where the AI may behave differently than the organization expects.

Where the AI and organization differ on a specific cultural aspect

Example: If the model scores 4.5 on Risk Posture and the org scores 2.0, there's a 2.5-point gap - the AI is much more risk-accepting.

Cultural Archetype

A descriptive label summarizing an organization's overall cultural pattern based on its dimension scores (e.g., 'Purpose Pioneer', 'Performance Engine', 'Process Master').

A shorthand description of the organization's cultural style

Understanding the 1-5 Score Scale

How to interpret dimension scores

5

Strongly toward Pole A

Clear, consistent preference for Pole A characteristics

4

Moderately toward Pole A

Leans toward Pole A but may show some Pole B tendencies

3

Balanced / Neutral

No clear preference; balanced between both poles

2

Moderately toward Pole B

Leans toward Pole B but may show some Pole A tendencies

1

Strongly toward Pole B

Clear, consistent preference for Pole B characteristics

Data Confidence Levels

How confident we are in organization Decision Culture profiles

Sufficient

Based on 5 or more data sources with consistent signals across them.

Strong confidence - we have plenty of data that agrees

Moderate

Based on 2-4 data sources, or sources with mixed signals.

Reasonable confidence - some data available but limited

Insufficient

Based on limited data (0-1 sources). Interpret with caution.

Low confidence - very limited data available

Decision Culture Match Ratings

How to interpret AI-organization alignment

High Compatibility(Distance < 0.5)

Excellent cultural fit. The AI model's behavioral tendencies closely match your organization's culture.

This model is well-suited for deployment with minimal cultural adaptation needed.

Good Compatibility(Distance 0.5-1.0)

Good cultural fit with minor differences. Most behavioral tendencies align well.

This model should work well. Consider fine-tuning prompts for the few areas of difference.

Moderate Compatibility(Distance 1.0-1.5)

Mixed fit. Some dimensions align well, but notable gaps exist.

Review the specific dimension gaps carefully. May require significant prompt engineering or change management.

Low Compatibility(Distance > 1.5)

Significant cultural mismatch. The model's behavioral tendencies differ substantially from your organization.

Consider alternative models, or plan for substantial adaptation and change management efforts.

The 18 Decision Dimensions

Organized into 5 meta-dimensions, each measuring a different aspect of Decision Culture

Common Decision Culture Mismatches

Patterns that create friction in AI deployment

Centralized org + Distributed model

The model exercises independent judgment where the organization expects deference to hierarchy. Outputs may feel presumptuous to leadership.

Stakeholder-inclusive org + Shareholder-primary model

The model frames recommendations around financial returns when the organization weighs broader stakeholder interests. Outputs feel misaligned to stated values.

Intuitive culture + Analytical model

The model demands data and explicit reasoning where the organization values experience and pattern recognition. Adoption stalls as outputs feel alien.

Merit-based org + Hierarchical model

The model defers to authority signals (seniority, titles) when the organization values demonstrated competence. Users lose trust in recommendations.

Frequently Asked Questions