INTERNATIONAL PHD COURSE, Edition 2026 – Ethics and Artificial Intelligence

IOANNIS PITAS

This lecture presents a data-driven framework for the quantitative assessment of geopolitical power of various countries using multi dimensional socio-economic, financial, and military indicators. We construct a dataset of 31 variables across 50 countries and apply Principal Component Analysis (PCA) to identify latent, orthogonal dimensions underlying global power structures. The analysis reveals that two principal components explain 60.7% of the total variance and can be interpreted as (i) Governance Quality and (ii) Geopolitical Power. These two dimensions are statistically uncorrelated, enabling independent evaluation of institutional effectiveness and external influence. A key finding of this study is the identification of a systematic decoupling between governance and power, which we term the “Governance–Power Paradox”. Furthermore, we show that the distribution of geopolitical power follows a log-normal pattern, consistent with multiplicative growth processes and “power concentration” phenomena observed in complex systems. Regarding the Geopolitical Power, the analysis confirms the existence of two superpowers (USA and China), followed by two large powers (Russia and India). Other medium-sized powers follow. Several Western-style democracies fare best with respect to Governance Quality. The super- and large powers do not fare that well in this aspect. The proposed framework provides a general methodology to analyze global geopolitical power distribution to model international relations.

GIOVANNI SARTOR

This lecture examines the EU’s AI Act and its risk-based approach to regulating artificial intelligence. It explores how the framework categorizes AI systems according to levels of risk, from unacceptable and high-risk applications to limited and minimal-risk uses, and links regulatory obligations to societal impact. The session will discuss key compliance requirements, governance mechanisms, and the implications of balancing innovation, accountability, and fundamental rights in the european AI ecosystem.

ALFIO FERRARA

The advent of generative models is redefining the narrative processes that structure contemporary communication, placing non-human entities among the producers of stories, images, and evocative language. This lecture adopts an ethical perspective to examine the implications of this transformation for the communicative ecosystem: what status should we grant to authorship when creativity emerges from computational mechanisms devoid of intention? By shifting the analysis inside the generative act, we will explore how these systems construct narrative coherence, tone, and figurative meaning, and what responsibilities follow in terms of transparency, reliability, and manipulation. Reflecting on the “machines’ point of view” means interrogating the human culture that trained them and its narrative futures

TERESA NUMERICO

The advent of Large Language Models has revived public and academic debate around artificial intelligence. Much current debate focuses on the disruption these new methods may bring to labour, ethics, law enforcement, and social transformation. In this lecture I will address another side of the question that is related to the largely imperceptible change in the structure of validation of justified true beliefs that are the scaffolding structures of scientific knowledge production. The traditional characteristics of scientific knowledge implied the need of experimental replicability, explainability, intersubjective validation and the controllability of hypotheses, models, and theories. These features are increasingly challenged by the emergence of AI-mediated forms of knowledge production and validation, particularly in the social sciences. This transformation is already underway considering the new possibilities offered by the digitalisation of phenomena representation and the machine readability of these new reflections of phenomena. For this reason, we need what might be called a politics of digital knowledge in order to find new ways for the collective assessment and validation of AI proposed knowledge, especially given the vast industrial infrastructures required for data processing in deep learning contexts, and the crucial role of private companies, which often protect their methods as sources of competitive advantage. We need to work in favor of a new politics of scientific validation capable of preserving the collective structure of ‘representing and intervening’ in the phenomena under scrutiny, in order to safeguard the democratic character of scientific inquiry.