System Function
Cassandra is an Information Relationship Management (IRM) system that allows users to extract highly specified and complex data from large corpuses of unstructured records. The system executes source-anchored reasoning pipelines, ensuring every output remains traceable to primary evidence. The system’s unique ability to reason over technical source material while maintaining full auditability makes it an ideal resource for both researchers and scholars.
The system does not independently recruit participants, interact with human subjects, influence subject behavior, or make autonomous determinations affecting participant rights, welfare, eligibility, or study outcomes.
Data Scope and Inputs
The system processes only data explicitly provided by the research team and only for approved research purposes. It does not scrape external sources, introduce new personal data, or perform secondary data use beyond protocol scope. Use of de-identified, coded, or limited datasets is supported where required.
Retrieval-Bounded Generation and Traceability
GraphRAG architecture grounds generated outputs in retrieved source materials through a structured graph-based retrieval layer. Outputs are traceable to their originating documents, enabling verification, citation, and auditability. This design mitigates unsupported or speculative generation and supports transparency and reproducibility.
AI Safeguards and Human Oversight
Automated outputs are advisory and non-determinative. All interpretations and conclusions require human review and remain the responsibility of the research team. The system does not replace investigator judgment or IRB oversight.
Privacy and Confidentiality
Confidentiality protections include role-based access controls, logical data segregation, and configurable restrictions on data visibility and export. Research data is not sold, shared, or used to train generalized AI models outside the approved research context.
Data Security
Administrative, technical, and physical safeguards are implemented, including encryption in transit and at rest, access logging, and authentication controls, consistent with commonly accepted institutional and industry standards.
Participant Interaction
The system does not communicate with research participants, obtain consent, provide interventions, or otherwise engage directly with human subjects.
Data Retention and Disposal
Data retention periods are configurable to align with institutional policy and approved protocols. Secure deletion mechanisms are available upon study completion or termination of access.
IRB Alignment
The software is intended for use only within IRB-approved activities. Any material changes to data type, scope, or analytical use that could affect human subjects protections require prior IRB review.
IRB Classification Note
The described GraphRAG system functions as an analytical research support tool and does not independently constitute human subjects research under 45 CFR 46.