Masterarbeit
Explainable AI for FAIRness recommendations in research software
Completion
in progress
Research Area
Intelligent Information Management
Students
Khushi Rajesh Zope
Advisers
Dr. Sheeba Samuel
Description
Current FAIR assessment approaches for research software have several limitations. First, they mainly provide scores or checklist-based evaluations without actionable guidance, making it difficult for developers to understand how to improve their software. Second, FAIR principles are often high-level and abstract, which makes translating them into concrete implementation steps challenging. Third, existing tools typically lack explainable and metric-linked recommendations that clearly justify why a certain improvement is needed. To address these problems, the thesis proposes to build a system that analyzes research software repositories using a selected set of FAIR4RS-based metrics and generates human-readable, actionable recommendations. To ensure the recommendations are accurate and grounded, the thesis plan to combine rule-based FAIR4RS checks with large language models using retrieval-augmented generation (RAG). The goal is to produce recommendations that are clear, precise, and explicitly linked to FAIR metrics, and potentially include small code or documentation templates to support implementation.
