Disciplinary Improvements: FAIROS-compliant agent AI systems for biodiversity research

Generative artificial intelligence (AI) is rapidly transforming how people find, analyze, and use information. In biodiversity science, where research depends on integrating data from thousands of specimens, publications, and observations, these tools hold enormous promise but also pose new challenges. Scientific progress, environmental resilience, and biosecurity all depend on trustworthy, accessible, and reusable data. Despite significant investments in biodiversity data infrastructure, many researchers still struggle to navigate complex data repositories, apply FAIR (Findable, Accessible, Interoperable, and Reusable) practices consistently, or take full advantage of available digital resources. This project advances the national AI research priority by ensuring that AI-powered research systems produce results that are FAIR, while following practices of Open Science, hence FAIROS. The project enhances established national and NSF cyberinfrastructure with the agent-based research platform iChatBio so that AI-assisted scientific workflows seamlessly comply with FAIROS principles. By lowering technical barriers to high-quality data use, the project expands participation in biodiversity research, supports education and workforce development, strengthens open science practices, and provides a model for responsible AI integration across scientific disciplines. The project designs, implements, and evaluates a suite of interoperable software agents that enable FAIROS-compliant biodiversity research within iChatBio. Four integrated tasks guide the work: (1) developing agents that leverage existing FAIR mechanisms in major biodiversity repositories and standards; (2) creating agents that mitigate gaps where FAIROS mechanisms are incomplete or absent, including automated data harmonization and identifier strategies; (3) implementing management and validation agents that generate machine-actionable workflow archives capturing inputs, outputs, provenance, metadata, and persistent identifiers; and (4) evaluating effectiveness, usability, and performance through research workflow templates, user surveys, and usage analytics. The project employs a co-development model in which computer scientists, biodiversity informaticians, and domain researchers collaboratively design reusable agent templates, validation strategies, and workflow templates. Outcomes include deployable FAIROS agents, standardized workflow-archive structures, documented design patterns, and evaluated best practices for AI-enabled research. These contributions establish a scalable framework for making agent-based scientific workflows reproducible, transparent, and reusable in biodiversity science and beyond. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Directorate for Biological Sciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2531902 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Elizabeth Ellwood | Institution: University of Florida, GAINESVILLE, FL | Award Amount: $598,751 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2531902 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2531902.html