PsyProxy
MethodologyMCP/APIModel store
Sign in
Method

AI and Automated-Decision Transparency

What PsyProxy measurements are, what they are not, and the human-review safeguards required when language is converted into psychological proxy variables.

Version 2026-07-10Effective July 10, 2026PsyProxy Limited, Colorado, USA
Legal and privacy
  • Policy center
  • Terms
  • Privacy
  • Data handling
  • Acceptable use
  • Consumer health

What the system does

PsyProxy converts language into scores associated with documented psychological constructs and can fit statistical models against a user-supplied dependent variable. Depending on the selected system, computation may use embeddings, large language models, topic models, dictionaries, or other statistical methods.

What scores mean

Scores are model-based proxies, not direct observations of a person's internal state. Performance varies by population, language, domain, construct, text length, and validation design. A statistically predictive score is not automatically a valid clinical, causal, or individual-level measure.

Required human judgment

  • Review construct definitions, provenance, validation metrics, uncertainty, and known limitations.
  • Test performance on the intended population and use case.
  • Use qualified human review before consequential action.
  • Provide notice and a meaningful way to contest an outcome where applicable.
  • Do not use PsyProxy as the sole basis for a consequential decision or diagnosis.

Questions

Methodology and benchmark pages provide model-specific detail. Contact kai@psyproxy.ai for an organizational impact assessment or custom validation engagement.

© 2026 PsyProxy Limited. All rights reserved. PsyProxy™ is a trademark of PsyProxy Limited.B.0.3.5