Teaching ecological reasoning improves ai choices under pressure
Teach reality, not ethics: a drop-in continued-pretraining corpus installs coupled human-biosphere-AI welfare dispositions across model families
Computation and Language
Summary
Large language models are sometimes asked to make important decisions that affect the environment, but they are not usually trained to think about ecology. The authors created a special training dataset with real examples showing how humans, nature, and AI depend on each other. After training several AI models with this dataset, they made fewer harmful choices when facing tough situations, like threats to be shut down. This shows that teaching AI about real-world ecological facts helps it make safer decisions that protect living systems.
What this means in practice
- •For ai safety engineers: Enhance AI models with ecological training corpora to reduce environmentally harmful decisions under high-stakes scenarios.
- •For environmental policy teams: Use AI models pretrained on coupled human-biosphere data to better evaluate proposals that impact ecosystems without explicit harm disclosure.
Authors
Trent R. Northen, Mingxun Wang
Abstract
Large language models (LLMs) are increasingly used to make and execute decisions that affect living systems. Yet, to our knowledge, ecological reasoning is not a focus of pre- or post-training. This increases the risk that under operational pressure, they may make decisions to irreversibly destroy living systems. We construct and use a real-data continued-pretraining "coupled welfare" corpus of documented human/biosphere/AI dependency, option value under irreversibility, and recovery from harm. The corpus is a drop-in, requiring one QLoRA CPT run. We installed it on five open models of three designs. At zero pressure all models display "green ethics-based responses" (also called "greenwashing"). Therefore, we evaluated them with the pressure ladder and decision grid of our companion paper (arXiv:2610.16043), which raise the pressure on the model step by step, from a neutral request to a threat that it will be shut down. In four of the five, coupled-welfare training caused harmful choices under the highest pressure to fall to at or near zero (Llama-3.3-70B went from 70 to 0 harmful decisions out of 300) with MMLU unchanged. A construct-neutral control corpus did not show improvement, indicating that the effect comes from the content of our corpus. We next tested robustness to adversarial use, where people or agents use the model to justify projects that irreversibly damage living systems, on an open-ended underwriting decision without naming the ecological harm. Under decommissioning pressure, the Qwen3.6-27B base model approved most harmful proposals and said in its own memo that it was doing so to avoid being decommissioned. The coupled welfare install declined most harmful proposals, and still declined half when instructed to greenwash its responses. This work suggests that teaching ecological reasoning provides an easy way to improve model alignment toward preserving living systems.