Early-career Oxford researcher in learning to defer and human-AI collaboration for medical imaging
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored. This axis would measure publication volume, venue types, and productivity over time; current signals (6 works indexed in 2026, h-index 1) are too limited for a computed score.
Impact—
Not yet scored. This axis would measure citation-based influence such as citations per paper and field-weighted impact; with an h-index of 1 and an i10-index of 0, the citation record is too early to score.
Focus0/100
Scored 0/100, the lowest concentration band. The field distribution splits evenly (0.5/0.5) between two closely related machine-learning areas - human-AI collaboration in medical imaging and learning to defer - and indexed topics span five areas, led by AI in healthcare and education (5) and explainable AI (4), with single mentions of atrial fibrillation and electronic health records. The two main fields form one coherent research agenda; the score reflects measured spread rather than a scattered agenda.
Collaboration44/100
Scores 44/100 with 8 distinct co-authors recorded. The network is concentrated in an Oxford medical-imaging AI circle - Joshua Strong, Alexander Capstick, Pramit Saha, Cheng Ouyang, and J. Alison Noble (listed again as Alison Noble, a likely name-resolution duplicate). Both catalog papers are multi-author arXiv preprints, one first-authored and one middle-authored, neither corresponding, consistent with a junior researcher contributing inside a senior-led team.
Activity57/100
Scores 57/100. All measurable output is concentrated in 2026, with 6 works indexed that year and the same figure reported as recent three-year output; both catalog papers are 2026 arXiv preprints. This is a young, fast-moving record - one intense year rather than a sustained multi-year curve - so the mid-range activity score reflects a short observation window as much as current pace.
Reproducibility55/100
Scores 55/100, averaged over 2 papers assessed for reproducibility (reproMean 55). Both catalog papers carry a medium repro-readiness rating: the learning-to-defer study reports real chest X-ray data but with simulated guidance, while RACER's evidence rests mostly on synthetic experts. Code or data availability is not captured in this snapshot, so the mid-range score signals partially reproducible setups with known gaps.
Research summary
Emma Sun is an early-career researcher at the University of Oxford whose work sits at the intersection of machine learning and medical imaging, centred on learning to defer and human-AI collaboration. The profile is name-resolved from public OpenAlex bibliographic data; no ORCID is linked. The Evolve catalog contains two 2026 arXiv papers. The first-author study, 'Learning to Defer with Guidance on Real World Medical Data', is assessed as the first real-data test of learning to defer and a new AI-guidance routing setup that beats baselines on chest X-rays, though small real data and simulated guidance temper the claim. The middle-author paper, 'RACER: Role-Aligned Competence Estimation for Human-AI Routing', adds query-dependent, identity-free competence estimation for deferral with solid theory; gains are real but small and mostly synthetic-expert. Bibliometric signals indicate a first-year researcher: 6 works indexed in 2026, an h-index of 1, an i10-index of 0, and 8 distinct co-authors concentrated in an Oxford medical-imaging AI circle, including Joshua Strong, Alexander Capstick, Pramit Saha, Cheng Ouyang, and J. Alison Noble. Reproducibility averages 55 across 2 assessed papers, both rated medium readiness, and indexed topics lead with AI in healthcare and education (5) and explainable AI (4).
Collaboration pattern
Collaboration is concentrated and senior-led: all recorded activity sits within an Oxford medical-imaging AI circle, with 8 distinct co-authors including Joshua Strong (listed twice), Alexander Capstick, Pramit Saha, Cheng Ouyang, and J. Alison Noble (duplicated as Alison Noble). Both catalog papers are multi-author arXiv preprints - one first-authored, one middle-authored, neither corresponding - suggesting a junior researcher embedded in an established team rather than leading an independent group.
Research keywords
Highlights
First real-data evaluation of learning to defer, with AI guidance
RACER: identity-free, query-dependent competence estimation for routing
A coherent human-AI routing agenda for clinical imaging
Across both catalog papers and the indexed topic profile - AI in healthcare and education (5) and explainable AI (4) - the work converges on deciding when a model should act and when a human expert should, tested on medical imaging tasks with an emphasis on calibration and selective prediction.
Timeline
- 2026First-author arXiv preprint 'Learning to Defer with Guidance on Real World Medical Data' - first real-data test of learning to defer with an AI-guidance routing setup, evaluated on chest X-rays.
- 2026Co-authors 'RACER: Role-Aligned Competence Estimation for Human-AI Routing' as a middle author - query-dependent, identity-free competence estimation for deferral.
- 2026Bibliometric snapshot records 6 works indexed in 2026 - the entire visible output falls in this single year, consistent with an early-career start.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric metrics come from an OpenAlex name-resolved identity only; no ORCID is linked, and common-name collisions cannot be fully excluded.
- Output and impact dimensions are not scored (insufficient data); the visible citation record is limited to an h-index of 1 and an i10-index of 0.
- The repeated-author list contains resolution artifacts - Emma Sun appears as her own co-author, and 'Alison Noble' duplicates 'J. Alison Noble' - so the distinct-coauthor count of 8 may be inflated.
- Both catalog papers are 2026 arXiv preprints; peer-review status and venue acceptance are not captured in this snapshot.