NCEAS researchers publish ten rules for AI coding in environmental science

Researchers at the University of California, Santa Barbara’s National Center for Ecological Analysis and Synthesis (NCEAS) have published a comprehensive framework to standardize the use of generative AI in environmental research. Released in PLOS Computational Biology, the guide addresses the rapid integration of AI coding tools into scientific workflows, which researchers noted outpaced institutional oversight and created inconsistent practices across teams. The initiative originated in 2023 as internal protocol for the Wildfire Resilience Index project, a multi-jurisdictional effort combining satellite imagery, land-cover data, and socioeconomic variables using Python and R pipelines. Lead researcher Rachel King noted that the team repeatedly encountered identical challenges regarding AI trust, access management, and session continuity. Rather than allow each of the dozens of NCEAS teams to develop independent workarounds, leadership convened a cross-disciplinary group of 22 researchers, developers, and data analysts to formalize best practices. The resulting publication, Ten simple rules for effective use of generative AI for code development in environmental science, fills a documented gap in existing literature by tailoring guidelines specifically to the messy, multi-source nature of environmental data and the varying technical proficiency within research teams. The ten rules are structured around three operational phases: project preparation and tool selection, active coding assistance, and post-development verification and documentation. The framework emphasizes transparent validation of AI-generated code rather than blind adoption. Co-author Cat Fong stressed that the guidelines deliberately avoid uncritical promotion of generative AI, instead focusing on practical tradeoffs relevant to small academic teams. Beyond technical methodology, the briefing highlights systemic risks accompanying AI adoption in science. Researchers warn that productivity gains are disproportionately benefiting male academics, while subscription models threaten to exclude underfunded institutions and researchers in low-income nations. The publication also notes that while generative AI usage exceeds 60 percent in some wealthy countries, adoption in developing regions remains near 5 percent, risking a deepening digital divide. Environmental infrastructure costs present another concern, with data centers projected to consume between 4 and 12 percent of U.S. electricity by 2030 and billions of gallons of water annually by 2028. The NCEAS team stops short of mandating AI use, arguing that implementation decisions require separate ethical review. Their central conclusion is that effective AI integration in environmental science is not an automatic advantage but a specialized competency. The field currently lacks structured training to help researchers navigate tool selection, prompt engineering, and code validation responsibly. By establishing a standardized, phase-based approach, the NCEAS guidelines aim to transform generative AI from a disruptive variable into a rigorously managed component of ecological and computational research.