Researcher at Texas State University building trustworthy, auditable ML for mental health NLP and road safety
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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: the available data covers only two reviewed 2026 arXiv preprints, with no publication counts or venue record. This axis would measure total scholarly output, venue placement, and annual publication volume once a fuller bibliometric record is available.
Impact—
Not yet scored: citation counts, h-index-style indicators, and downstream usage are unavailable for this early-stage record, whose two preprints are recent. This axis would measure citations, uptake of the released benchmarks, and influence on practice once citation data accumulates.
Focus0/100
The focus axis scores 0, reflecting an even 50/50 split in the raw field distribution between transportation safety AI (applied machine learning) and computational social media / mental health NLP. Both directions are pursued in parallel across the two 2026 preprints, so no dominant field has emerged yet; consolidation would lift this score.
Collaboration32/100
Evolve scores this axis at 32, based on a small co-author network: 4 distinct coauthors appear across the two reviewed papers. Subasish Das recurs in the frequent-coauthor list, alongside Anika Baitullah and Anandi Dutta, indicating a compact, repeat-partner working circle rather than a broad international network; expanding independent collaborations would raise this score.
Activity—
Not yet scored: no year-by-year publication stream is available beyond the two 2026 preprints, so research cadence cannot be computed. This axis would measure annual publication and preprint activity, showing whether output is sustained or concentrated in a single burst.
Reproducibility38/100
Reproducibility scores 38, computed as a mean of 38 across the 2 papers with reproducibility reviews. The Bluesky mental-health benchmark paper is rated low readiness, while the dementia-related crash severity paper reaches medium readiness; Evolve notes honest statistics in the latter but a single train/test split, weak absolute performance, and unevaluated deferred cases, and noisy mental-health labels (κ=0.222) in the former.
Research summary
Gaurab Chhetri is a researcher at Texas State University whose reviewed work centers on making applied machine learning trustworthy in two socially consequential domains. According to Evolve's catalog, both of the researcher's 2026 arXiv preprints are first-authored (non-corresponding): a validation-aware weak-supervision framework for building mental health benchmarks on Bluesky, and a selective-deferral framework for auditable, calibrated prediction of dementia-related crash severity. The raw field distribution splits evenly (0.5/0.5) between computational social media / mental health NLP and transportation safety AI, and a shared methodological thread — candid label audits, leakage control, calibration, and explicit human-review handoffs — runs across both papers. Evolve's reviews are measured rather than glowing: suicidal-ideation labels validate fairly well while mental health labels are noisy (κ=0.222), and the crash-severity work is credited for honest statistics but criticized for a single train/test split, weak absolute performance, and unevaluated deferred cases. Reproducibility averages 38 across 2 reviewed papers (low and medium readiness), collaboration scores 32 with 4 distinct coauthors — Subasish Das listed twice as the anchor partner — while output, impact, and activity metrics are not yet available.
Collaboration pattern
The reviewed work is produced within a compact team: 4 distinct coauthors appear across the two 2026 first-authored preprints. Subasish Das is listed twice in the frequent-coauthor list, making them the anchor collaborator, with Anika Baitullah and Anandi Dutta also recurring. This pattern suggests small, repeat-partner teams spanning both research directions — transportation safety AI and mental health NLP — rather than large multi-institution consortia.
Research keywords
Highlights
Weak-supervision benchmark building on Bluesky, with candid label audits
Selective deferral for auditable dementia-related crash severity prediction
One trustworthy-AI playbook across two domains
Across both 2026 preprints the same quality-control toolkit recurs — label validation audits, outcome leakage control, calibration, and explicit deferral to human review — applied evenly across a 50/50 field split between mental health NLP and transportation safety AI.
Timeline
- 2026First-authored arXiv preprint: Building Trustworthy Mental Health Benchmarks on Bluesky — a validation-aware weak-supervision framework with candid label audits.
- 2026First-authored arXiv preprint: auditable, calibrated AI for dementia-related crash severity prediction using a selective deferral framework to support human review.
- 2026Both reviewed preprints appear in the same year with first authorship (non-corresponding), splitting output evenly between transportation safety AI and mental health NLP.
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 signals are limited to Evolve's radar on the 2 reviewed papers; output, impact, and activity scores are missing, and no citation data is included.
- This profile is name-resolved from public bibliographic data and the provided ORCID (0009-0000-0124-4814); name ambiguity cannot be fully excluded.
- Both papers are 2026 arXiv preprints; peer-review status and final venues are not reflected in this data.
- Radar scores (collaboration 32, focus 0, reproducibility 38) are computed from a very small sample of 2 papers and should be read as provisional.