Support to present accepted PRISM AI Safety Research Fellowship work on trust calibration in healthcare AI at NeurIPS 2026 in Atlanta.
Through the PRISM AI Safety Research Fellowship, under the mentorship of Prof. Zeenath Reza Khan, my team and I are studying trust calibration in healthcare AI under incomplete information.
The project asks when users should rely on, verify, delay, question, or reject AI recommendations, especially when AI outputs include explanations, confidence cues, uncertainty cues, or warning/verification prompts. This work has produced an accepted NeurIPS 2026 workshop paper, a public HAI-TRUST dataset on Zenodo, a second paper accepted for oral presentation at ICAMAC 2026, and an empirical study with survey design completed and data collection in progress.
I am an equal-contributing author on the accepted NeurIPS workshop paper and am seeking support to travel to Atlanta to present the work in person.
Project Links:
OpenReview paper: https://openreview.net/forum?id=48s9v6mYez
Zenodo dataset: https://zenodo.org/records/22855853
The immediate goal is to present our accepted NeurIPS 2026 workshop paper on trust calibration in healthcare AI and use feedback from the NeurIPS community to refine and strengthen the accepted paper, the empirical paper in progress, and the HAI-TRUST manuscript.
The overall research goal is to improve how AI systems are evaluated in high-stakes human decision settings. In healthcare AI, the safety problem is not simply whether people trust AI more or less, but whether their reliance is calibrated to the quality, limits, and uncertainty of the AI output.Our work studies this through a thematic review, conceptual framework, HAI-TRUST scenario-bank dataset, empirical study, and a methodological paper on scenario-bank design, construction, and expert review.
I will achieve this by attending the NeurIPS 2026 Workshop on Human-AI Coevolution in Atlanta, presenting the work in person, discussing it with researchers in AI safety, human-AI interaction, healthcare AI, and evaluation, and bringing that feedback back into our ongoing empirical work.
This funding would support my travel to Atlanta for NeurIPS 2026 to present our accepted workshop paper in person.
Minimum funding requested: $1,000. This would cover accommodation.
Maximum funding goal: $2,300. This would cover the full estimated travel cost:
$1,000 accommodation
$550 round-trip travel to Atlanta
$400 meals during travel and workshop attendance
$250 local transportation
$100 poster printing/presentation materials
Registration is not included in this request because PRISM will cover conference registration for one author. I have also applied/will apply to other travel funding sources, including BlueDot Rapid Grants, but no travel grant or funding is confirmed yet.
Our research team is part of the PRISM AI Safety Research Fellowship, under the mentorship of Prof. Zeenath Reza Khan. The team consists of four research fellows, including Kumari Neha Priya (myself), Favour Chisom Akabogu, Rokhaya Diagne, and Tustee Mazumdar.
I am an equal-contributing author on our accepted NeurIPS 2026 workshop paper and have actively contributed to strengthening the project through all phases, including literature synthesis, conceptual framework development, ethics application preparation, scenario-bank development, manuscript preparation, and empirical study design. I am also the team member seeking funding to travel to Atlanta and present our work in person.
Our team has successfully produced an accepted NeurIPS workshop paper, a publicly released HAI-TRUST dataset on Zenodo, and a second paper accepted for oral presentation at ICAMAC 2026. We are also working on an empirical paper.
My research background combines AI safety, healthcare AI, machine learning, empirical analysis, and AI governance. I hold an MS in Business Analytics (4.0 GPA) from the University of Illinois Urbana-Champaign, where I currently work as an academic researcher at Gies College of Business. My ongoing work includes research on emerging U.S. healthcare policy, technology, and innovation, developing a healthcare research website for teaching and research, and co-authoring research on responsible AI in public health with Prof. Sridhar Seshadri.
Previously, I was a SPAR Research Fellow under the mentorship of Joël N. Christoph (Harvard Kennedy School), studying market-based compute permits for frontier AI governance. I have also contributed to an international AI healthcare regulation white paper, worked on LLM evaluation through Handshake AI, Turing, Mercor, and Micro1, and served as a reviewer for two NeurIPS 2026 workshops revieweing 6 papers. I completed BlueDot Impact's AGI Strategy course and received a full conference registration scholarship for Stanford AI+HEALTH 2026.
Relevant work includes:
NeurIPS accepted paper: https://openreview.net/forum?id=48s9v6mYez
HAI-TRUST dataset: https://zenodo.org/records/22855853
SPAR research report: https://library.sparai.org/reports/market-based-compute-permits-for-frontier-ai-safety-vr1nia/
CV: https://drive.google.com/file/d/1ETvXsFKOrHWqppKb3dxfcTSwCyMFNaI-/view?usp=sharing
The project itself will continue, but I may not be able to attend NeurIPS in person because the travel cost is high for me to cover personally. If I cannot travel, my team may need to rely on lower-cost alternatives, but we would lose the opportunity to present the work directly to the NeurIPS community.
The impact would be losing in-person feedback, collaboration, networking, and visibility at one of the world’s leading AI/ML conferences. That feedback is important now to further refine and strengthen our accepted paper, empirical paper, and HAI-TRUST manuscript.
Personally, attending NeurIPS would also be meaningful for my long-term PhD and research career goals combining AI safety, healthcare AI, human-AI evaluation, and AI governance.
I have not raised external travel funding for this NeurIPS trip yet. PRISM is expected to cover conference registration for one author, but this does not cover my travel, accommodation, meals, local transportation, or other related costs. I have applied/will apply to other travel funding sources, including BlueDot Rapid Grants, but nothing is confirmed yet.
Comments
Kumari Neha Priya
10 days agoHi @NeelNanda, I've seen your support for AI safety researchers and projects on Manifund, and I was hoping you might consider supporting mine as well.
I'm travelling to NeurIPS 2026 in Atlanta to present our accepted research on trust calibration in healthcare AI under incomplete information. PRISM has covered my conference registration and flights, and I'm raising funds for the remaining expenses.
I'd be very grateful if you could consider contributing, even a small amount, to help me present our work and receive feedback from the research community. Thank you for considering it!
Kumari Neha Priya
10 days agoUpdate: PRISM has also covered my flight expenses, so the funding goal is now revised to $1,950 to reflect the remaining estimated costs, including accommodation, meals, local transportation, and poster printing.
Kumari Neha Priya
10 days agoOur team consists of fellows from countries underrepresented in AI safety research, where opportunities to attend international conferences are limited and the cost of travelling abroad is often unaffordable. We've all put so much time and effort into this project, and PRISM has already supported us as much as they could.
I'm hoping to present our work at NeurIPS, get feedback that can help us improve our research, and connect with researchers who could open up future opportunities for our team. This project means a lot to all of us, and any support towards my attendance would also be supporting the work and efforts of my teammates.
We would genuinely appreciate any contribution, no matter how small. Thanks to everyone who takes the time to read about our work and considers supporting us!
Kumari Neha Priya
10 days agoHi @RyanKidd ,
I'm familiar with your work at MATS Research (I meant who doesn't know the CEO and founder!) and have seen you support various AI safety projects through Manifund, including researchers' travel to NeurIPS.
I would be grateful if you could take a look at my project and consider supporting it. Any contribution, regardless of the amount, would be greatly appreciated and would help cover my remaining conference travel and accommodation expenses.
Thank you for your time and consideration!
Kumari Neha Priya
10 days agoHi @keri,
I came across your contribution towards an undergraduate researcher's participation at ACML 2026 and thought of reaching out.
I would be grateful if you could take a look at my project and consider supporting it. Any contribution, regardless of the amount, would be greatly appreciated and would help cover my remaining travel and accommodation expenses for NeurIPS 2026.
Thank you for your time and consideration!
Kumari Neha Priya
about 1 month ago@gnutella Thanks so much!
gnutella
about 1 month ago@KNPriya When will you hear from the BlueDot Rapid Grants? Congrats and good luck!
Kumari Neha Priya
about 1 month ago@gnutella Thanks! I heard back from BlueDot, but unfortunately, my application wasn’t selected for funding. I’m very grateful for your generous support, which will make a real difference in helping me attend NeurIPS!
Kumari Neha Priya
about 1 month agoHi @Austin, my proposal has been flagged as AI-generated by Pangram and consequently excluded from default project listings, reducing its visibility to potential funders. As someone with experience in LLM evaluation, I find this concerning. Structured and professional writing does not automatically mean AI slop.
I also tested this AI detector's accuracy. When I first submitted my proposal, it was flagged as 86% AI-generated. I then kept editing my content and removing more lines, yet the score jumped to 90%. This raises serious concerns about the detector's accuracy and how it is unnecessarily penalizing my proposal. I should not have to compromise the quality of my writing just to get past this flag.
I completely understand this is done to filter low-quality submissions, but a detection score alone should not determine a proposal's credibility or visibility. Could you please review this flag and reconsider the restriction?