Research

Local clinical knowledge for AI

With Gillings Innovation Lab support, we study how structured feedback from clinicians can improve AI-generated cases. Daktari-DX is the interface clinicians use to review those cases and contribute local clinical knowledge.

Active research. Gillings Innovation Lab (GIL) is the funder; Daktari-DX is the clinician-facing interface. No model has demonstrated clinical effectiveness. GIL, RLEF Data Generation, and Daktari-DX are related but not interchangeable names.

AIM AHEAD and AugMed

AugMed is a platform for running clinical case-review experiments. Our AIM-AHEAD colorectal cancer risk-assessment study examines how AI recommendations affect clinicians' judgments, screening recommendations, confidence, and attention to patient information.

Analysis. No proven improvement, sample sizes, effect estimates, publication status, or named health-system partner until a paper is out.

Clinical AI Evaluation Methods

Developing rigorous methodologies to assess AI-assisted clinical decision support systems, from algorithm performance to real-world clinical impact.

Clinician training in Ethiopia

Our developing research agenda examines how case-based practice, feedback, and information presentation affect clinicians' reasoning, confidence calibration, and learning. It is intended to inform CollectiveGood's Ethiopia training effort and future evaluations with local partners.

Developing. No executed training study, learning gains, accredited CME, public platform, or verified deployment. No platform link. Partner names come from the approved lab flyer; the MOU's terms are not described here.

Healthy Future 2.0

We study how AI-informed recommendations and family preferences can help prioritize maternal and child health home-visit content in rural China. CareLoom is our open-source platform for scheduling visits and delivering curricula.

In development. No enrollment, comparative results, or Kenya implementation claims.

Collective intelligence for curbside consultation

Our NIH NIMHD-funded Phase I STTR project develops and evaluates asynchronous peer consultation for rural clinicians. The research examines rapid skill assessment, selection of relevant reviewers, and aggregation of independent judgments, alongside feasibility and workflow fit. Clinical effectiveness remains to be established.

Phase I, newly awarded. Never state diagnostic superiority, completed clinical validation, or a broadly available clinical service.

Patient preferences and diabetes prevention

With Humana Foundation support, we study how patient preferences can inform text-message nudges for Diabetes Prevention Program participation. Formative research and preference-survey development are guiding preparation for a feasibility pilot.

Formative work and pilot preparation. No efficacy claim; no learned personalized or adaptive policy.

Measuring HIV care quality in Vietnam with standardized patients

Unannounced standardized patients measure the quality of HIV care across providers in Vietnam, with a focus on how stigma shapes what patients receive, plus prescribing-pattern analysis from medication photos. Builds on R34MH121251 (MPI Sylvia and Smith, 2019 to 2022), which developed the standardized-patient stigma measures in China.

Active data collection. Not among the flyer's seven selected projects; retained here as substantiated funded work.

Information Markets for community health

We study whether incentives and market mechanisms can aggregate local observations into useful public health information. A published Ghana pilot examined the feasibility and acceptability of collective intelligence-based COVID-19 surveillance. The broader agenda concerns how to elicit, combine, and evaluate community knowledge.

Research program with one published pilot. Kenya extension paused since 13 May 2026; Ghana pilot published. No outbreak-prediction accuracy, representativeness, or active Kenya trial claims.