![[ Visit Client Website ]](images/banner.gif)
The first paper uses LLM-based agents grounded in survey data to simulate responses to family-support policies related to fertility intentions. By comparing simulated outcomes to observed data, the authors assess the feasibility of this approach as a lower-cost, scalable tool for early-stage policy pretesting. This work demonstrates how AI can support more efficient and evidence-informed policy design, with implications for strengthening policy responsiveness and promoting the common good.
The second paper employs persona-based LLM agents to replicate published health preference findings, testing the extent to which synthetic data can approximate population-level patterns. Findings suggest these approaches may be useful for survey piloting, scenario testing, and early-stage research, particularly in resource-constrained settings. At the same time, the study highlights important limitations, reinforcing the continued necessity of human expertise and lived experience in producing valid and equitable research outcomes.
The third paper evaluates a multi-agent AI system designed to support literature screening in systematic reviews. By comparing AI-assisted screening to traditional team-based approaches, the authors demonstrate how such systems can improve efficiency while preserving human judgment, transparency, and methodological rigor. This work positions AI as a supportive tool that can expand access to evidence synthesis capacity, particularly for teams with limited resources.
The fourth paper presents a scoping review of AI-based dietary intake measurement, a foundational method for nutrition research and food policy. The authors identify critical equity gaps in validation and representation, noting that many AI tools have not been adequately tested among populations most affected by measurement bias. These findings underscore the importance of ensuring algorithmic fairness and inclusive design so that advances in measurement technology do not reinforce existing disparities.
Taken together, these papers illustrate how AI can be leveraged to generate, test, and synthesize knowledge in ways that are both innovative and accountable. Across diverse applications - policy simulation, preference modeling, evidence synthesis, and measurement systems - these studies demonstrate the potential of human-centered AI to expand research capacity while maintaining rigor and ethical responsibility. This symposium highlights how inclusive, methodologically rigorous inquiry can advance social work research by strengthening evidence-informed decision-making, expanding access to research tools, and promoting more equitable outcomes for the common good.