Title: What Does AI Think About Sustainability? Resource URL: https://www.wbs.ac.uk/news/what-ai-think-sustainability/ Publication Date: 2024-07-08 Format Type: Article Reading Time: 8 minutes Contributors: Isabel Fischer; Source: Warwick Business School Keywords: [Artificial Intelligence, Sustainability, AI Ethics, ESG Strategy, AI in Climate Policy] Job Profiles: Chief Sustainability Officer (CSO);Policymaker;Academic/Researcher;Sustainability Task Forces;Sustainability Manager; Synopsis: In this article, Warwick Business School professor Isabel Fischer analyzes research comparing AI-generated sustainability priorities with human perspectives and finds that while some generative AI models show alignment, they are not yet reliable for making nuanced sustainability decisions. Takeaways: [AI has potential in sustainability applications, but it does not yet fully replicate human reasoning on complex environmental and social issues., Research comparing AI responses to human stakeholder views found varying levels of alignment., AI can identify broad sustainability trends, but lacks consistency in ranking priorities across different sectors., Trust and ethical concerns, such as bias and stakeholder representation, must be addressed before AI can be a reliable sustainability advisor., AI-human collaboration will be key to future sustainability efforts, balancing technological efficiency with human judgment and ethical considerations.] Summary: Artificial intelligence is being explored as a tool for sustainability, offering potential solutions in areas such as environmental monitoring, climate-conscious investing, and resource conservation. However, AI’s ability to "think" about sustainability remains questionable, as human perspectives on the issue are complex and often divided. Research conducted by Isabel Fischer and her colleagues analyzed how generative AI models—such as ChatGPT, Gemini, Copilot, and Claude—prioritize sustainability issues compared to human stakeholders from sectors including business, public policy, NGOs, and education. Findings showed that AI models varied significantly in their alignment with human perspectives. Gemini demonstrated the strongest correlation with human stakeholders, particularly in the public sector, while other models such as Claude and Copilot showed inconsistencies. AI models often failed to accurately rank lower-priority sustainability issues, and their reasoning lacked nuance compared to human judgment. The study highlights that while AI can process vast amounts of data, it struggles to grasp the deeper ethical, social, and political aspects of sustainability. Trust remains a significant concern, as AI-generated insights might misrepresent stakeholder priorities or fail to capture the full scope of sustainability challenges. While AI can assist with sustainability efforts, it is not yet capable of replacing human reasoning in this domain. Future advancements in AI ethics, training, and contextual awareness may improve its reliability, but for now, human leadership remains crucial in driving meaningful sustainability progress. Content: ## Introduction ### AI and Sustainability Consider an environment in which artificial intelligence (AI) collaborates with human stakeholders to navigate the complexities of sustainability and propose viable solutions. Advancements in AI ecosystems—such as Google’s Gemini—and the exponential growth of data have created unprecedented opportunities. Yet, a fundamental question endures: Can AI genuinely comprehend the multifaceted nature of sustainability and align its analyses with human values and priorities? ## Historical Evolution of Sustainability ### Origins and Early Definitions The term “sustainability” was first introduced at the United Nations’ inaugural Environmental Conference in Stockholm, 1972. That milestone catalyzed global discourse on environmental stewardship. ### From Millennium Declaration to SDGs In 2000, the United Nations Millennium Declaration established comprehensive objectives to eradicate poverty, hunger, disease and illiteracy, while simultaneously addressing environmental degradation and gender inequality. These aims laid the groundwork for the subsequent Sustainable Development Goals (SDGs)—a more ambitious agenda adopted in 2015, targeting progress by 2030. ## Divergent Perspectives on Sustainability Despite these high-level commitments, political measures have often fallen short of the scale required to combat climate change. Concurrent challenges—rising sea levels, volatile food and energy prices and looming economic downturns—underscore the vulnerability of global systems under population pressure. Effective achievement of SDG targets demands collaboration among public agencies, private enterprises, nonprofits and academic institutions. However, research undertaken in partnership between two European business schools revealed a critical barrier: stakeholder groups frequently hold divergent interpretations of what constitutes sustainability. A 2022 analysis of Google search queries identified “What is sustainability?” as the most common sustainability-related question, highlighting widespread conceptual confusion. Without a shared understanding, multi-​stakeholder initiatives will struggle to align on objectives and metrics. ## Comparative Analysis of Human and AI Priorities ### Research Methodology To assess AI’s ability to mirror human sustainability concerns, researchers in India invited representatives from private companies, public agencies, nonprofit organizations and educational institutions to rank 23 sustainability concepts. These concepts ranged from the interconnection of environmental, economic and social issues to specific sustainability indicators. The same ranking task was then issued as prompts to four major generative AI (GenAI) chatbots—ChatGPT, Google’s Gemini, Microsoft’s Copilot and Anthropic’s Claude—and, in one instance, another popular model (Bard). ### Findings by Sector #### Business and Private Sector • Top Human Priorities: Climate change; gender equality; education. • AI Alignment: – Copilot and Claude closely echoed human rankings, also placing climate change first. – ChatGPT emphasized the interconnection of environmental, economic and social dimensions. – Gemini prioritized the needs of current versus future generations. • Correlation with Human Rankings: – Gemini: 0.734 (highest alignment) – Claude: 0.54 (lowest among the four) #### Public Sector • Top Human Priorities: Water security; healthy ecosystems; climate change. • AI Alignment: – Gemini again aligned most closely, ranking climate change highest. – Other chatbots failed to prioritize these concerns accurately, nor did they correctly rank lower-priority items such as green chemistry, poverty, the concept of interconnection or generational equity. • Correlation with Human Rankings: – Gemini: 0.56 (highest alignment) – Copilot: 0.31 (lowest alignment) #### Nonprofit Sector (NGOs) • Top Human Priorities: Healthy ecosystems; environmental interconnection; gender equality; responsible consumption. • AI Alignment: – Each chatbot—excluding Gemini—identified at least one of these priorities, yet inconsistencies persisted. – ChatGPT, Bard and Copilot correctly designated the ‘triple bottom line’ as a lower priority. • Correlation with Human Rankings: – ChatGPT: 0.76 (highest alignment) – Claude: 0.13 (lowest alignment) #### Education Sector • Top Human Priorities: Water security; quality education; population growth management. • AI Alignment: – All models highlighted education and system interconnection, reflecting partial agreement with human stakeholders. • Correlation with Human Rankings: – Claude: 0.75 (highest alignment) – Copilot: –0.22 (notable divergence) ## Implications for AI in Sustainability Discourse Although generative AI excels at reproducing textual patterns, replicating the depth and nuance of human reasoning remains an ongoing challenge. Trust, risk assessment and broader ethical considerations—such as data privacy—are central to any deployment of AI in sustainability contexts. Current evidence indicates that AI cannot yet be relied upon to generate recommendations that fairly represent diverse stakeholder viewpoints. Human judgment continues to be indispensable, particularly for ensuring that marginalized perspectives are neither overlooked nor misunderstood. ## Conclusion The intersection of AI and sustainability presents promising avenues for innovation—from environmental monitoring to climate-aware investment tools. Yet, to harness these capabilities responsibly, AI systems must be refined to align more closely with the plurality of human values and priorities. The discourse is far from settled. Achieving a sustainable future will require an ongoing partnership: the analytical power of AI, guided and enriched by human wisdom.