Grass, flowers, and other greenery growing across a robotic brain to show AI thinking about sustainability

Green thinking: Research shows AI cannot yet replicate human thinking and make trustworthy suggestions on sustainability

Imagine a world where AI sits alongside us, grappling with the complexities of sustainability and helping to identify solutions.

With advanced AI technology like Gemini (an AI ecosystem from Google) and a tidal wave of data flooding our digital landscape, the possibilities seem endless.

However, amidst this whirlwind of innovation, a key question remains. Can AI ‘grasp’ the complexities of sustainability, aligning with our human values and perspectives?

The term “sustainability” was officially coined at the inaugural Environmental Conference held by the United Nations in Stockholm in 1972. It was a defining moment in the global discussion of environmental issues.

Since then, the scope of sustainability has evolved and expanded. In 2000, the United Nations Millennium Declaration created a united front against poverty, hunger, disease, illiteracy, environmental degradation, and gender inequalities. It also outlined social and environmental objectives for member nations, laying the foundation for change.

These were succeeded by the Sustainable Development Goals (SDGs), a set of even more ambitious targets to achieve by 2030 that reflect the growing global challenges we face.

Why are views on sustainability so divided?

Despite these noble aspirations, the journey has been fraught with challenges. Political responses have often fallen short of addressing the urgency and scale of climate change. At the same time, rising sea levels, food, and energy prices, combined with a global recession, have exposed the vulnerabilities inherent in supporting the world’s burgeoning population.

The complexity of sustainability issues means that achieving the targets set out in the SDGs requires collaboration among a diverse range of stakeholders.

However, my research with Lutz Preuss, of Kedge Business School, and Bimal Arora, of Aston Business School, has revealed a fundamental weakness in this framework: the divergence of perspectives held by different groups.

The fact that a 2022 study identified “What is sustainability?” as the most frequently searched sustainability-related phrase on Google, illustrates how many people are struggling to grasp the fundamental concept of sustainability.

This lack of shared understanding poses a significant obstacle to achieving common goals. The success of any initiative involving multiple stakeholders will depend on how those different groups make sense of the sustainability issue at hand. If those perspectives do not align, realising sustainable development objectives becomes increasingly challenging.

Can generative AI models reflect human views on sustainability?

Drawing on our research, we compared the sustainability priorities among stakeholders in India with those expressed by generative AI (GenAI).

Technologies like SoraOpenAI in entertainment and large language models (LLMs) used for research and education have already demonstrated AI's ability to streamline processes and enhance productivity.

The same technology could offer innovative solutions to pressing ecological concerns, from environmental monitoring to resource conservation and climate conscious investing.

For example, Rho AI was developing an ESG investment tool to make it easier for small-scale investors to understand and integreate in formation on companies’ climate impact into their decisions about where to invest. I have analysed that project in an earlier article.

But before we entrust AI with a seat in sustainability discussions, we must understand its compatibility with human perspectives. Can AI truly reason about sustainability issues with the same depth and nuance as humans?

In India, stakeholders from various sectors including private, public, NGO and education industries participated in a survey to rank 23 sustainability concepts. These ranged from the interconnection of environmental, economic and social issues to sustainability indicators.

The same questions were used as prompts for four prominent GenAI ChatBots: ChatGPT, Gemini, Copilot, and Claude.

This comparative analysis sheds light on how AI systems perceive and prioritise sustainability issues compared to human stakeholders.

How do human and AI priorities on sustainability differ?

Within ‘human participants’ of the business and private sectors, we found that climate change, gender equality, and education are top priorities.

There was a surprising degree of alignment between these stakeholder priorities and the rankings by Copilot and Claude, highlighting climate change as the most pressing issue.

However, ChatGPT had a different perspective, emphasising the interconnection of environmental, economic, and social issues.

Meanwhile, Gemini stood out by placing a strong emphasis on the needs of current versus future generations.

As for low priorities, ChatGPT correctly identified the ‘triple bottom line’ as a lower priority.

Gemini showed the highest correlation with human stakeholder priorities (0.734), indicating a significant and positive alignment.

Conversely, Claude had the lowest correlation (?0.54), suggesting a notable disparity in its rankings.

How consistent is GenAI at reflecting human views?

Similarly, we observed a consensus in the public sector on crucial sustainability concerns such as water, healthy ecosystems, and climate change.

Once more, Gemini proved to be aligned with human reasoning, emphasising climate change as the top priority for this group.

However, the other GenAI Chatbots failed to identify this priority accurately, and none of them correctly ranked lower factors such as green chemistry, poverty, the interconnection of environmental factors, and the needs of current versus future generations.

This discrepancy underscores the necessity for further refinement in AI reasoning.

Gemini exhibited the highest correlation with human stakeholder priorities (0.56), indicating a positive alignment. Conversely, Copilot demonstrated the lowest correlation (?0.31), suggesting a notable disparity in its rankings.

NGOs displayed distinct priorities, focusing on healthy ecosystems, environmental interconnection, gender equality, and consumption.

Which AI chatbots mirror human views most closely?

While each of the chatbots, excluding Gemini, identified one of these priorities, their rankings were inconsistent. ChatGPT, Bard and Copilot correctly identified the lowest factor as the ‘Triple Bottom Line’. Yet discrepancies persisted, indicating the need to improve AI's understanding of stakeholder perspectives.

ChatGPT demonstrated the highest correlation with human stakeholder priorities (0.76), suggesting a solid alignment. Conversely, Claude exhibited the lowest correlation (0.13), highlighting areas for enhancement in its reasoning capabilities.

Stakeholders in the education sector prioritised water, education, and population growth. Interestingly, GenAI Chatbots also emphasised education and environmental interconnection, indicating a partial alignment with human priorities.

Among the Chatbots, Claude demonstrated the highest correlation with human stakeholder priorities (0.75), suggesting a solid alignment in reasoning.

Conversely, Copilot exhibited the lowest correlation (-0.22), indicating a mismatch with human perspectives and highlighting areas for improvement in its reasoning capabilities.

Can we trust AI to make sensible suggestions on sustainability?

While GenAI tools are good at replicating patterns to generate text, fully replicating human reasoning is a different ball game. The concepts of trust and balancing risk are key when using GenAI (along with other ethical issues, such as data privacy).

Can we trust AI to make sensible suggestions on sustainability that clearly consider and represent different stakeholder positions? Based on our findings, the answer is no, not yet.

For the time being at least, the value of human reasoning is irreplaceable, especially when it comes to diverse perspectives that are often overlooked by widely available information.

However, in the future AI might be further refined to have the capability to align more closely with diverse human values and priorities, steering us towards a future where sustainability thrives.

One thing is sure: the debate about AI and sustainability is far from over. Our path towards a sustainable future hinges on the fusion of human wisdom and collaboration with AI.

This article was originally published in IT Now magazine by the Chartered Institute for IT.

Further reading

Working on the jagged frontier: How companies should use generative AI

Pass the IP: How generative AI will re-shape intellectual property

How can the world reach net zero?

How do stakeholder groups make sense of sustainability?


Isabel Fischer is Reader in Information Systems at Warwick Business School. She teaches Creating Digital Communities on the MSc Management of Information Systems and Digital Innovation; and Introduction to Consulting, Developing Consulting Expertise, and International Management on the MSc Management at Warwick Business School.

Learn more about how to harness the power of AI for your organisation on the four-day Executive Education course Business Impacts of Artificial Intelligence at WBS London at The Shard.

For more articles on Sustainability sign up to our Core Insights newsletter.