Matt Murtagh

From 01/01/2025 to 10/06/2025

Challenges

Making International Development Research Findable, Accessible, Interoperable and Reusable (FAIR)

Pennsylvania State University

Matt Murtagh is a researcher from Trinity College Dublin studying a PhD in Artificial Intelligence. He is researching methods to improve systematic review of international development evaluation data using Large Language Models and Knowledge Graphs. He has an undergraduate degree in Economics and Politics from Trinity College Dublin and an MSc. in Economics for Development from the University of Oxford. Previous to the NGI Fellowship, he was selected for an ODI Fellowship in Rwanda, where he worked at the National Institute of Statistics Rwanda (NISR) to establish the institute’s first data science team.

International development interventions are increasingly reliant on rigorous, evidence-based evaluations to shape policy and practice. However, the volume of available evaluation data is growing exponentially, making it difficult for researchers, policymakers, and practitioners to identify, organize, and integrate new findings efficiently. Despite initiatives like the International Initiative for Impact Evaluation (3IE)’s Development Evidence Portal, much of the evaluation landscape remains fragmented, inconsistently formatted, and time-consuming to parse.

InterDev addresses this challenge by applying advanced AI and semantic web technologies to improve how development evaluation research is curated and shared. The goal is to develop tools and frameworks that make this data FAIR – Findable, Accessible, Interoperable, and Reusable – enabling more timely, data-informed decision-making across international development stakeholders.

This project emerges at a critical juncture, as the international development sector grapples with both data overload, a shift toward open, interoperable research practices and a global funding crisis brought on by the defunding of USAID. By embedding AI within open-access workflows and linking social science rigor with computational methods, InterDev aims to close the gap between policy aspirations and data realities and allow existing data to be more efficiently reused for novel projects.

The NGI Enrichers Fellowship at Penn State supports the refinement of the platform and introduces valuable collaborations with interdisciplinary researchers in social and organizational informatics. The host institution’s emphasis on human-centered, ethically reflective technology design has provided a rigorous environment to critically test and improve InterDev.

During the first quarter of the fellowship, a full round of user testing was conducted with stakeholders in the international development research community. Feedback gathered focused on usability, interpretability of outputs, and alignment with organizational workflows. This has directly informed the design of the next iteration of the UI/UX and backend data pipeline. In addition to this research I have also collaborated with many local researchers at the university – I am currently completing research on enhancing scientific claim verification models by integrating LLMs with knowledge graphs, on inducing behaviour economic biases in LLMs, producing a dataset on fake news in less resourced languages and comparing user behaviour patterns between LLM and static interfaces.

Research Publications and Scientific Validation

Three collaborative papers are currently in preparation for submission to leading AI conferences, including AAAI and NeurIPS. These papers are co-authored with researchers from both Europe and the United States, demonstrating active transatlantic collaboration. Topics include LLM integration with knowledge graphs for scientific claim verification, behavioral bias modeling in LLMs, and comparative analysis of user behavior across interactive interfaces.

Prototype Testing and Iterative Improvement of InterDev

The fellowship facilitated a comprehensive round of user testing for the InterDev platform, generating actionable feedback on the usability and interpretability of the system. This user-centered process has directly influenced the platform’s next development phase and contributed to a clearer pathway for future implementation and adoption.

Strengthened Transatlantic Research Collaborations

The fellowship has enabled meaningful exchanges between my host lab at Penn State and my home lab, ADAPT at Trinity College Dublin. These connections have opened avenues for future joint grant applications, student mobility, and mirrored research projects exploring ethical and human-centered AI in global development.

Innovation Development and Planned Patent Submission

A patent application is in preparation for submission by the end of the fellowship. The patent outlines a novel “Digital Library Assistant” system that semantically enhances search and discovery of public domain literature. This is a direct product of fellowship-supported work combining LLMs with structured knowledge frameworks.