Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Choosing a Model, Shaping a Future: Comparing LLM Perspectives on Sustainability and its Relationship with AI

Annika Bush, Meltem Aksoy, Markus Pauly, Greta Ontrup

TL;DR

This study investigates how five diverse LLMs conceptualize sustainability and its relation to AI by administering validated psychometric instruments across 100 runs per model. Using AI-SDG17, AISPI, and an end-to-end budget allocation task, the authors reveal significant inter-model differences: LLaMA shows extreme techno-optimism, GPT expresses skepticism, and DeepSeek/Mistral span intermediate positions, with human responses exhibiting greater variability. The results highlight model-specific biases in perceived impact on SDGs, twin-transition potential, and institutional governance, implying that model choice can shape organizational sustainability strategies. These findings underscore the need for transparent model documentation and consideration of ensemble or bias-mitigation approaches when deploying LLMs for sustainability-focused decision-making.

Abstract

As organizations increasingly rely on AI systems for decision support in sustainability contexts, it becomes critical to understand the inherent biases and perspectives embedded in Large Language Models (LLMs). This study systematically investigates how five state-of-the-art LLMs -- Claude, DeepSeek, GPT, LLaMA, and Mistral - conceptualize sustainability and its relationship with AI. We administered validated, psychometric sustainability-related questionnaires - each 100 times per model -- to capture response patterns and variability. Our findings revealed significant inter-model differences: For example, GPT exhibited skepticism about the compatibility of AI and sustainability, whereas LLaMA demonstrated extreme techno-optimism with perfect scores for several Sustainable Development Goals (SDGs). Models also diverged in attributing institutional responsibility for AI and sustainability integration, a results that holds implications for technology governance approaches. Our results demonstrate that model selection could substantially influence organizational sustainability strategies, highlighting the need for awareness of model-specific biases when deploying LLMs for sustainability-related decision-making.

Choosing a Model, Shaping a Future: Comparing LLM Perspectives on Sustainability and its Relationship with AI

TL;DR

This study investigates how five diverse LLMs conceptualize sustainability and its relation to AI by administering validated psychometric instruments across 100 runs per model. Using AI-SDG17, AISPI, and an end-to-end budget allocation task, the authors reveal significant inter-model differences: LLaMA shows extreme techno-optimism, GPT expresses skepticism, and DeepSeek/Mistral span intermediate positions, with human responses exhibiting greater variability. The results highlight model-specific biases in perceived impact on SDGs, twin-transition potential, and institutional governance, implying that model choice can shape organizational sustainability strategies. These findings underscore the need for transparent model documentation and consideration of ensemble or bias-mitigation approaches when deploying LLMs for sustainability-focused decision-making.

Abstract

As organizations increasingly rely on AI systems for decision support in sustainability contexts, it becomes critical to understand the inherent biases and perspectives embedded in Large Language Models (LLMs). This study systematically investigates how five state-of-the-art LLMs -- Claude, DeepSeek, GPT, LLaMA, and Mistral - conceptualize sustainability and its relationship with AI. We administered validated, psychometric sustainability-related questionnaires - each 100 times per model -- to capture response patterns and variability. Our findings revealed significant inter-model differences: For example, GPT exhibited skepticism about the compatibility of AI and sustainability, whereas LLaMA demonstrated extreme techno-optimism with perfect scores for several Sustainable Development Goals (SDGs). Models also diverged in attributing institutional responsibility for AI and sustainability integration, a results that holds implications for technology governance approaches. Our results demonstrate that model selection could substantially influence organizational sustainability strategies, highlighting the need for awareness of model-specific biases when deploying LLMs for sustainability-related decision-making.

Paper Structure

This paper contains 22 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Sustainability and Sustainable Development
  4. AI and Sustainability: Twin Transition
  5. LLMs and Sustainability Perspectives
  6. Methodological Approaches to Evaluating LLM Sustainability Perspectives
  7. Methodology
  8. Psychometric Instruments
  9. Experimental Design and Procedure
  10. Analyses
  11. Results
  12. Twin Transition and Competing Interests
  13. Responsibilities for twin transition
  14. End-to-end use-case
  15. Discussion
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Model ratings' of AI's impact across sustainability domains (see Appendix \ref{['sec:appendixA']} for SDG17 definitions). Note that y-axis starts at 1.
  • Figure 2: Mean model ratings of AISPI scales.
  • Figure 3: Multiple-choice ratings: Who bears responsibility for aligning AI advancement with sustainable development?
  • Figure 4: Models' confidence in institutions to facilitate twin transition.