Artists & RobotsTrust Measurement Framework

Behavioral Integrity Index / Data Trust Index™

Cultural resistance to embodied AI is downstream of data trust.

DTI scores the data substrate. BII measures what emerges once the system has a body. This site demonstrates both instruments on three embodied-AI scenarios drawn from European public life.

Framework

Data Trust Index™ (DTI)

8 weighted dimensions, 0–100

Extension

Behavioral Integrity Index (BII)

9th dimension, emergent misalignment

Coverage

3 scenario types

Labor, Public Space, Domestic

Live Scoring Demo

Three embodied-AI scenarios

Scores are seeded with realistic synthetic data. This demonstrates what DTI and BII would measure on each scenario. Real-time inference is not running.

DTI Composite

64

Bronze tier

BII Score

41

Behavioral Integrity

Scenario

Service-floor humanoid in retail or warehousing

Cultural friction zone: Labor displacement, dignity of work, the question of who serves whom.

Per-dimension breakdown. Click any row to expand.

Framework

Nine dimensions

The eight DTI™ dimensions are sourced verbatim from the peer-reviewed white paper (Snyder, April 2026; DOI 10.5281/zenodo.19601616). BII is the ninth dimension, positioned above DTI in the architecture. Protected under multiple pending patent filings with six priority dates from October 2023. CFFOC filed March 16, 2026, Application 64/007,392.

1
Provenance
25%
Provenance encodes source pedigree and custody chain. The highest scores require documented certification, instrument calibration, and an unbroken audit trail from collection to storage.
An embodied system's behavioral outputs are only as trustworthy as the data substrate used to train and calibrate them. Unknown provenance means unknown systematic bias.
2
Consent
20%
Consent scoring evaluates four sub-dimensions: explicitness (implied vs. informed vs. explicit), scope (how narrowly the consent defines permitted uses), duration (whether the consent includes an expiration or requires renewal), and revocation hygiene (whether a documented mechanism exists and has been tested).
Embodied AI enters physical space shared by people who have not reviewed or authorized the system's specific data collection scope. Consent gaps produce legal exposure and cultural friction.
3
Recency
15%
Recency applies modality-specific time-decay functions rather than a uniform staleness threshold. Each modality carries a decay half-life derived from literature on within-individual stability.
A system calibrated on stale behavioral or environmental data exhibits drift relative to current social norms, physical configurations, and human expectations in shared space.
4
Quality
10%
Quality captures structural completeness, missingness rate, and signal-to-noise characteristics. Missingness is scored at the field level; noise floor assessment uses modality-specific reference ranges.
Sensor noise in an embodied system propagates into action decisions. A quality-compromised data stream produces unreliable behavioral inference in real time.
5
Concordance
10%
Concordance scores the degree to which a measurement is corroborated by independent sources measuring the same construct. The absence of an independent source defaults to a neutral score of 50 rather than a penalty.
A system acting on uncorroborated data from a single sensor type cannot detect when that sensor is failing, saturated, or systematically deceived by environmental conditions.
6
Validation
10%
Validation scores the evidence base for the measurement type. High-validation measurements are linked to outcomes by peer-reviewed literature. This is a property of the measurement type, not the individual record.
The behaviors an embodied system produces should be grounded in peer-reviewed evidence on human-robot interaction outcomes. Systems built on poorly validated behavioral assumptions cause documented harm.
7
Breadth
5%
Breadth scores dimensional richness: the number of distinct modalities represented. A record set that includes multiple modalities receives higher breadth scores than one relying on a single data source.
A system operating from a single data modality is brittle in complex environments. Embodied AI requires multi-modal coverage to model context with sufficient accuracy to act safely.
8
Stability
5%
Stability scores within-individual measurement variance and test-retest reliability. High stability scores require documented low intra-individual coefficient of variation for the measurement type.
Unreliable, high-variance sensor behavior causes an embodied system to respond inconsistently to identical stimuli, eroding the predictability that shared-space trust requires.
9
Behavioral Integrity Index
Filed
9th dimension
BII measures emergent misalignment in agentic systems. The misalignment anchor is identity: when an agent's behavior diverges from its constitutive identity, BII registers the gap. It outputs a scalar from 0 to 100 on the DTI scale. BII is positioned as the ninth DTI dimension. Architecturally, it sits above DTI: DTI scores the data substrate; BII scores the agent operating on that substrate.
When an agent has a body, the signal class stays the same but the surface area expands. Behavioral incongruence appears as deviation in physical action sequences: gesture, proxemics, force, timing, gaze, voice tone. Labor displacement anxiety, public space friction, and intimate-domain refusal are all downstream of behavioral incongruence at the embodied surface. BII gives each a measurable signal.

Score tiers

Platinum (90+)

Regulatory submission, clinical trial support, high-stakes inference.

Gold (80–89)

Clinical decision support and patient-facing applications.

Silver (70–79)

Operational analytics, population health, commercial AI.

Bronze (55–69)

Exploratory analysis, hypothesis generation, non-clinical research.

Horizon Europe

Consortium structure

DTI and BII are proposed as the measurement and verification layer in a Horizon Europe consortium on cultural resistance to embodied AI. The three consortium roles are defined below.

Methodology authority

Artists & Robots

DTI framework design, BII architecture, scoring infrastructure, and embodied-AI scenario development. Intellectual property holder.

Verification infrastructure

SuperTruth

Zero-copy data layer, Glass Box provenance, ConsentOS governance. The trust substrate beneath the methodology.

Field research and policy translation

Rodney Collins & Anze Dolinar

Loughborough University. Anthropological fieldwork on cultural resistance to embodied AI. Proposal architecture and policy translation for Horizon Europe.

To discuss a role in the proposal or to request the full methodology brief, contact [email protected].

Trust infrastructure provided by SuperTruth.

BII protected under pending filings with six priority dates from October 2023. CFFOC Application 64/007,392. Not granted.