THE ROLE OF GREEN-AI IN SUSTAINABLE PLANNING: AN EXPLORATION ON THE CASE OF TARANTO, ITALY

This work explores the potential of generative artificial intelligence as a strategic and reflective Green AI, understood not as an energy-efficient technology, but as a high-analog processing environment capable of analyzing, critically evaluating, and reformulating sustainability strategies. Through a proof-of-concept on the case of Taranto, a city that symbolizes the contradictions between industrial development and environmental degradation, the ability of large-scale linguistic models (LLMs) to generate—when appropriately trained—elaborations comparable to the products of a reflective agent, integrating long-term quantitative and qualitative analyses, is tested. The model was guided through a multi-temporal reasoning exercise. Taking the perspective of a 1950s scientist, it formulated development strategies based on the knowledge of the time, then reevaluated them from a contemporary perspective. In this simulation, the Green AI explained criteria, probabilistic weights, and motivations, exhibiting argumentative transparency despite the notorious opacity of the black-box problem. The results suggest that an LLM, while not a reflexive agent, can replicate some of its utilities in analytical and decision-making contexts. Thus, beyond optimizing technical parameters, Green AI can become a tool for strategic reflection on ecological transition and sustainable planning. However, the analysis does not hide its critical limitations related to documentary bias and the risk of amplifying inequalities, requiring ethical safeguards for responsible applications in policy contexts.