Researchers at UC Santa Barbara's National Center for Ecological Analysis and Synthesis (NCEAS) have published a new framework to help environmental scientists use generative AI for coding without sacrificing research rigor. The "ten simple rules" guide, released in PLOS Computational Biology, grew out of repeated internal debates about when to trust AI suggestions and how to document AI-assisted work - conversations happening in labs everywhere as AI tools reshape scientific workflows.
"It started as something we needed for ourselves," said ecologist and data scientist Rachel King. "We kept repeating the same conversation, project after project: whether to trust a suggestion, how much access to give an AI agent, what to do when a long chat session forgets a decision it made an hour earlier. At some point it made more sense to work it out together, once, than have every team rediscover it independently."
From one project to a community standard
The framework emerged from the team's work on the Wildfire Resilience Index, an open-access tool that measures how prepared communities and landscapes are for wildfire. The project required stitching together satellite imagery, land-cover data and socioeconomic variables across two countries and 13 jurisdictions - a task that began in 2023 with early AI tools and ended with a new generation of AI coding assistants that made earlier lessons obsolete.
That experience wasn't unique to the wildfire team. Around NCEAS, a think tank for environmental science, dozens of research teams were having the same conversations: junior researchers relying on AI they didn't fully understand, senior scientists skeptical it could be trusted, and everyone in between inventing their own rules for responsible use. The center brought together 22 researchers, developers and data analysts to hash out a shared approach through literature review and months of co-writing.
"What started as guidance for our own community ended up filling a gap nobody else had addressed," said senior author Cat Fong. "Existing advice was written for software engineers, or for science in the abstract. Almost none of it accounted for what our field actually looks like - messy, multi-source data, small teams, wildly different levels of coding experience in the same room. And because some of us were openly skeptical of GenAI going into this, what came out of it is honest about the tradeoffs, not just enthusiastic about the tool."
Ten rules for before, during and after coding
The rules are organized around three phases of working with AI: preparation before coding, best practices during coding, and verification and documentation after. The authors are direct about the risks of getting this wrong, pointing to uneven access to AI tools across the scientific community.
"GenAI's benefits are not landing evenly," King said. "Male researchers report larger productivity gains than their female counterparts, and a recent UN report found that while roughly two-thirds of people in some high-income countries use GenAI tools, usage in many low-income countries hovers near just 5%."
The researchers also flag that as leading AI companies shift toward paid tiers, the most capable tools risk becoming inaccessible to researchers at underfunded institutions. The infrastructure behind generative AI carries environmental costs, too: data centers are projected to consume 4-12% of all U.S. electricity by 2030 and up to 32 billion gallons of water per year by 2028. Fong also noted that AI's rise has coincided with rising unemployment among recent computer science graduates and a steep decline in software development job postings, "raising hard questions about who is responsible for a talent pipeline that public investment built and private industry has since disrupted."
The authors stop short of prescribing whether AI should be used for a given task, arguing that question deserves its own ethical scrutiny. Once a researcher has decided to use it, they argue, doing so well is a skill - one the field hasn't yet been given the tools to build. For those working in AI for Science & Research, the framework offers a concrete starting point.
Why this matters for science and research professionals
For researchers who write code, the practical takeaway is that responsible AI use requires deliberate process, not just good prompts. The NCEAS framework gives teams a shared vocabulary for decisions they're already making: how much access to grant an AI agent, how to verify generated code, and how to document what the AI did so the work remains reproducible. Teams that adopt these rules now will have an advantage in producing research that holds up to scrutiny - and in avoiding the costly rework that comes from trusting AI output without verification.
Your membership also unlocks:
