Pediatric surgeon by training. Meta-researcher by conviction.
I study how AI changes the way we do science — not just faster, but structurally differently. My work sits at the intersection of clinical research, AI systems, and philosophy of science.
The problem I named: When AI assists research, it introduces epistemic drift — a silent corruption of the reasoning chain that leaves no trace in final results. Peer review can't catch it. Reproducibility checks miss it. Current safeguards assume human-controlled reasoning. AI breaks that assumption.
The other problem I'm working on: Evidence hierarchies built around RCTs misfire on operator-dependent medicine (surgery, IR, endoscopy, anaesthesia). I'm building OPERA — a constraint-driven appraisal rubric that scores trials on what they actually controlled for, not what design label they wore.
What I'm building:
| Project | What it does |
|---|---|
| epistemic-drift | Taxonomy of how AI corrupts scientific reasoning — companion to the A-series papers |
| citecheck | Verify manuscript citations against PubMed, CrossRef, OpenAlex — trust but verify |
| sr-pipeline | End-to-end systematic review automation: dedup + AI screening + PRISMA — 68 tests ✓ |
I write about medical research and how to integrate AI in academic works at https://www.tuyentranmd.com/.
Core belief: AI assists thinking. You own the science.
- Accepted: Neonatal Anorectal Malformations — Journal of Pediatric Surgery (in press)
- Under review: "Artificial Intelligence and the Epistemic Drift of Scientific Research" — defining a new class of epistemic failure in AI-assisted research
- In flight: OPERA framework — conceptual paper + empirical companion
- In flight: Constitutional framework for AI in science — why research needs pre-commitment constraints, not just post-hoc guidelines
"The crisis is not that AI makes bad science — it's that AI makes science that looks indistinguishable from good science."