top of page

GOVERNANCE & RESPONSIBLE AI

Auditable measurement of AI impacts people.

Validated, traceable, independent of the model being assessed.

Responsible AI programs require evidence — not just monitoring. Receptiviti provides validated psycholinguistic measurement of how AI affects the people it interacts with: cognitive state, emotional trajectory, distress signals, rapport. Derived from language, grounded in peer-reviewed science, structured for documentation. The kind of evidence that holds up outside the organization.

THE MEASUREMENT GAP

Most AI oversight measures what the model does.
Not what it does to people.

Bias testing, hallucination rates, toxicity filters, and policy conformance measure model behavior. They don't measure the effect of the model on the person - whether the AI increased their cognitive load or reduced it, whether it escalated their emotional state or stabilized it, or whether it built trust or eroded it. This human-impact dimension is not instrumented in most AI governance stacks. Receptiviti is that instrumentation.

WHAT CURRENT TOOLS MEASURE

Model-side behavior

Toxicity, hallucination, bias, policy conformance, drift — all properties of the model's output. None of them measure what that output does to the person receiving it.

WHAT'S MISSING

Human-impact measurement

Cognitive load, emotional trajectory, distress signals, rapport — the dimensions that determine whether an AI interaction is beneficial, neutral, or harmful. Currently unmeasured in most stacks.

WHAT RECEPTIVITI ADDS

Validated, independent signal

200+ psycholinguistic dimensions derived from language — not inferred by a model, not produced by the system being assessed. Traceable to 34,000+ peer-reviewed citations.

WHERE IT FITS

Where it fits in the NIST AI Risk Management Framework.

The NIST AI Risk Management Framework defines four core functions: Govern, Map, Measure, Manage. The Measure function calls for continuous, quantitative assessment of AI risks — including impacts on individuals. Receptiviti fills the human-impact measurement slot: validated, structured, documentable scores of what AI interactions do to the psychological state of the people in them.

Govern

Policies, accountability, culture

Map

Risk identification, context

Measure

Quantify, analyze, assess risks

Manage

Prioritize, respond, monitor

NIST AI RMF - MEASURE FUNCTION

Bias & Fairness

Security

Human Impact

Explainability

The human-impact dimension of the Measure function asks: what does the AI actually do to the people it interacts with? That question requires measurement that is independent of the AI, grounded in validated science, and structured for documentation.

Cognitive load — does the AI increase or reduce the burden of the interaction?

Emotional trajectory — does emotional state escalate or stabilize over turns?

Distress signals — are vulnerability indicators rising across the relationship?

Rapport — is the AI building or eroding relational trust?

WE PUBLISH. WE CONTRIBUTE.

Research grounding the measurement.

The science behind Receptiviti's measurement is active, published, and peer-reviewed. These studies establish what psycholinguistic measurement can do - and what it can reliably predict.

PUBLISHED: PNAS Nexus, 2024

Large language models display human-like social desirability biases in personality surveys

Salecha, Ireland et al.,2024. Large language models display human-like response biases when they infer they are being evaluated, with effects up to 1.20 human SD across GPT-4, Claude 3, Llama 3, and PaLM-2. Co-authored by Molly Ireland (Receptiviti).

Read the paper →

PUBLISHED: Clinical Psychological Science, 2026

Replicability and validity of a new AI assessment of PTSD from patient language

Kjell, Ganesan, Boyd et al., 2026. AI-based psychological assessment from language, when grounded in validated measurement, produces replicable, clinically meaningful results across sequential evaluation with preregistered models. Co-authored by Ryan Boyd (UT Dallas), academic partner to Receptiviti.

Read the paper →

SCOPED

Interaction state as a missing dimension in AI evaluation
The case for human-state signals as a first-class eval criterion alongside accuracy, helpfulness, and harmlessness.

​Research partnerships →

FOR GOVERNANCE PLATFORMS

WORKING WITH GOVERNANCE PLATFORMS

Receptiviti as an embedded measurement layer.

Governance and compliance platforms — AI GRC tools, risk registries, audit frameworks — govern model risk, security, and policy conformance. Most don't measure the psychological or behavioral quality of AI interactions. Receptiviti provides that measurement as an embeddable signal: validated, science-based, independently produced, structured for documentation alongside existing governance artifacts.
 

If you're building or operating an AI governance platform and want to discuss adding human-impact measurement to your stack, we'd like to talk.

Receptiviti does not claim to satisfy specific regulatory requirements, which vary by jurisdiction and evolve rapidly. What it provides is the validated, independent, structured measurement that human-impact oversight requires - wherever those requirements land.

Auditable measurement of AI's human impact.

Validated · Independent · Traceable to peer-reviewed science · Structured for documentation

bottom of page