Research
Systematic reviews and active research frameworks at the intersection of clinical psychiatry and AI diagnostic safety.
HIV-Associated Kaposi Sarcoma: A Systematic Review
A comprehensive systematic review examining the clinical presentation, epidemiological patterns, and treatment outcomes of Kaposi Sarcoma in HIV-positive populations, with particular focus on Sub-Saharan African cohorts.
AI Safety in Psychiatry Diagnostic Frameworks
Active research workspace building safety constraint architectures for AI-assisted psychiatric diagnostics.
Diagnostic Taxonomy Mapping
Mapped psychiatric diagnostic categories (DSM-5-TR) to LLM output patterns. Identified edge cases where model reasoning diverges from clinical safety constraints.
Edge Case Identification
Cataloging diagnostic edge cases — contraindication blindness, history omission, context collapse. Building override taxonomy with severity classification.
Safety Constraint Architecture
Designing safety constraint frameworks that lock clinical guardrails into AI diagnostic pipelines before patient-facing output.
Validation Testing
Testing framework against real-world clinical scenarios. Measuring override accuracy, false positive rates, and safety compliance metrics.
Contraindication Override
Model recommends treatment without checking patient history for contraindications.
History Blindness
Model ignores documented patient history when generating recommendations.
Context Collapse
Model fails to integrate dietary, lifestyle, or social context into diagnostic reasoning.
Panel Incomplete
Model initiates treatment recommendations without required lab work or diagnostic panels.
The preprint is available on request while the review is under peer review.
REQUEST_PREPRINT