Marassi, Lidia (2025) Human-Centred Generative AI: Ethical Challenges, Social Impacts and Environmental Sustainability. [Tesi di dottorato]

Documento PDF marassi_lidia_38.pdf Visibile a [TBR] Amministratori dell'archivio Download (17MB) | Richiedi una copia

Abstract

This thesis explores the ethical, social, and environmental implications of Generative Artificial Intelligence (GenAI), situating its analysis within the broader paradigm of Human-Centred AI. After defining the conceptual foundations of this paradigm (Chapter I) rooted in human dignity, transparency, and sustainability — the research examines how generative systems, while enabling unprecedented forms of creativity and automation, raise new questions about responsibility, equity, and ecological impact (Chapter II). The study adopts a multidisciplinary perspective that bridges philosophy, ethics of technology, data governance, and sustainability studies. Through this lens, it articulates three interrelated macro areas that together define the scope of Generative AI’s societal impact. First, the environmental dimension addresses the material costs of computation, focusing on energy consumption CO2 emissions, and water use associated with large-scale model training. It examines how efficiency gains, while valuable, may produce rebound effects that increase the overall ecological footprint (Chapter III). Second, the social dimension explores the transformations in labour, creativity, and cultural production brought about by generative tools, highlighting both opportunities and risks for innovation, and the need for AI literacy (Chapter IV). Third, the governance dimension analyses the evolving relationship between ethics and regulation (Chapter V) from the anticipatory role of ethical reflection to the institutionalization of norms in practical frameworks. The final section (Chapter VI) proposes a synthesis, outlining possible pathways for implementing human-centred Generative AI through sustainable, transparent, and procedurally competent design practices. It suggests that ethical reflection must remain dynamic and adaptive, especially as new paradigms further blur the boundaries between assistance and autonomous agency (Chapter VII). Ultimately, the thesis argues that only through the continuous negotiation between ethical ideals, technical feasibility, and environmental responsibility can artificial intelligence evolve as a genuinely human-aligned and sustainable technology.

Downloads

Actions (login required)

Modifica documento