13. AI and Ecolinguistics: Building Ecosophies to Stop AI Amplifying Environmental Harm

How do we prevent AI from amplifying destructive environmental narratives at a massive scale - potentially 100 billion words per day?Mariana Roccia and Jorge Vallego, from the H4rmony Project, reveal how ecolinguistics and ecosophies can reshape how large language models engage with ecological issues whilst addressing cultural and linguistic bias in AI-generated environmental discourse.This conversation explores how mainstream LLMs celebrate Coca-Cola as a "cultural icon" or patio heaters as "brilliant" without acknowledging environmental costs unless explicitly challenged. The team shares how they developed Theophrastus, an open-source assistant built on ChatGPT, instructed with an ecosophy: a living framework of ecological values that guides language generation toward planetary well-being rather than profit.We discuss how word embeddings cluster dominant narratives together in multidimensional space, why fine-tuning and reinforcement learning can shift those embeddings toward ecologically aligned responses, and how system prompts embed ecosophy into every AI interaction. The team explains their approach using preference datasets rather than imposed answers, working with the International Ecolinguistics Association's 1,500+ researchers to ensure cultural and linguistic representation.Mariana discusses why language representation matters, explaining how AI models are predominantly trained in English, which risks amplifying cultural imbalances and losing local ecological knowledge that's vital for different cultures. Jorge explains why transparency around environmental ethics in AI matters as much as addressing carbon footprint, and why major AI players need to adopt ecosophies just as they address gender and racial bias.This episode continues our new short series featuring conversations from the ⁠⁠⁠Building Bridges: A Symposium on Human-AI Interaction⁠⁠⁠ held at the University of Warwick on 21 November 2025. The symposium was organised by ⁠⁠⁠Dr Yanyan Li⁠⁠⁠, Xianzhi Chen, and Kaiqi Yu, and jointly funded by the Institute of Advanced Study Conversations Scheme and the Doctoral College Networking Fund, with sponsorship from Warwick Students' Union.AI Ethics NowExploring the ethical dilemmas of AI in Higher Education and beyond.A University of Warwick IATL PodcastThis podcast series was developed by Dr Tom Ritchie and Dr Jennie Mills, the module leads of the ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠IATL module ⁠"The AI Revolution: Ethics, Technology, and Society"⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ at the University of Warwick. The AI Revolution module explores the history, current state, and potential futures of artificial intelligence, examining its profound impact on society, individuals, and the very definition of 'humanness.'This podcast was initially designed to provide a deeper dive into the key themes explored each week in class. We want to share the discussions we have had to help offer a broader, interdisciplinary perspective on the ethical and societal implications of artificial intelligence to a wider audience.Join each fortnight for new critical conversations on AI Ethics with local, national, and international experts.We will discuss:Ethical Dimensions of AI: Fairness, bias, transparency, and accountability.Societal Implications: How AI is transforming industries, economies, and our understanding of humanity.The Future of AI: Potential benefits, risks, and shaping a future where AI serves humanity.If you want to join the podcast as a guest, contact [email protected].