CLINICAL TRIAL LISTING — CANONIC Community Learning Study

inherits: hadleylab-canonic/IRBS

Axiom

ClinicalTrials.gov registration for the CANONIC Community Learning Study. Observational, retrospective, multi-arm. CaribChat (Caribbean) and MammoChat (US). Governed AI health navigation with structurally anonymized community learning ledger.

Brief Title

Community Learning Patterns in Governed AI Health Navigation

Official Title

Community Learning Patterns in Governed AI Health Navigation: A Retrospective Observational Study of Structurally Anonymized Ledger Data from Federated Cancer Navigation Services

Brief Summary

This study characterizes the community learning patterns that emerge when patients, caregivers, and clinicians interact with governed AI health navigation services that capture questions on a structurally anonymized, append-only ledger. The study analyzes questions asked across two federated navigation services: CaribChat (Caribbean cancer navigation across eight countries) and MammoChat (US breast health navigation). The governance architecture enforces anonymization at the point of capture; no personally identifiable information is collected at any stage. The study evaluates whether governance-native data architecture provides adequate human subjects protections in jurisdictions with and without research ethics legislation, and characterizes the compounding community intelligence that accumulates when navigation questions are ledgered and governed.

Detailed Description

Background

AI-driven health navigation services generate conversational data that constitutes a community learning resource when governed properly. The CANONIC governance framework provides structural anonymization (no PII fields in the data schema), append-only immutability (ledger entries cannot be modified or deleted), and cryptographic integrity verification. Each navigation service inherits identical governance constraints, enabling modular, federated study arms under a single protocol.

Study Population

Intervention

None. This is an observational study of existing, anonymized ledger data.

Outcomes

Primary: Characterization of community learning patterns (question taxonomy, temporal trends, geographic references) in governed AI health navigation.

Secondary: Cross-arm comparison of learning patterns between Caribbean and US populations; evaluation of governance-native data architecture as human subjects protection in jurisdictions without research ethics legislation; measurement of community intelligence compounding over time.

Study Design

Field Value
Study Type Observational
Observational Model Other (federated community learning ledger)
Time Perspective Retrospective
Number of Arms 2 (expandable via amendment)

Arms and Interventions

Arm Description Intervention
CaribChat (Arm A) Caribbean cancer navigation, 8 countries, 55+ sessions None (observational)
MammoChat (Arm B) US breast health navigation, Florida primary, 20+ sessions None (observational)

Eligibility

Field Value
Ages Eligible All ages
Sexes Eligible All
Accepts Healthy Volunteers Yes
Sampling Method Non-probability (all ledgered sessions included)

Inclusion: Any session ledgered on a governed TALK instance (CaribChat or MammoChat) during the study period.

Exclusion: None. All ledgered sessions are included. The data is structurally anonymized; there is no mechanism to exclude based on individual characteristics.

Contacts and Locations

Role Name Institution
Principal Investigator Dexter Hadley, MD/PhD CANONIC Foundation
Co-Investigator Marisa Nimrod, MD Trinidad and Tobago

Study Locations

Facility City Country
caribchat.ai (virtual) Port of Spain Trinidad and Tobago
mammochat.ai (virtual) Orlando United States
Role Organization
Sponsor CANONIC Foundation
Collaborator Caribbean Association of Oncology and Hematology (CAOH)

Keywords

community learning, governed AI, health navigation, cancer navigation, breast cancer, Caribbean, structural anonymization, append-only ledger, federated learning, CANONIC

*IRBS CARIBCHAT LISTING 2026-03-18*