At the Université Libre de Bruxelles, a Short Term Scientific Mission brought together Rossella Resi and the research teams of Prof. De Valeriola and Prof. De Brabanter to examine a growing linguistic phenomenon in the domain of Artificial Intelligence: the rise of anthropomorphized neologisms.

In public and specialized discourse alike, AI systems are increasingly described using terms traditionally associated with human cognition. We commonly state that a model “thinks,” “reasons,” or “understands.” Such expressions, while often metaphorical, implicitly attribute cognitive capacities to computational systems, capacities that are typically considered uniquely human. Since language shapes perception, this linguistic shift may influence how emerging technologies are conceptualized, evaluated, and integrated into society.

Metaphorical naming has long played a central role in scientific and technical development across disciplines. However, within the rapidly evolving field of AI, the intensity and frequency of anthropomorphic terminology appear to be increasing. This trend raises important questions about its cognitive, communicative, and societal implications.

During the STSM, the research teams initiated a collaborative investigation into whether this tendency, already identified in English, extends similarly to French, Dutch, and Italian. The project seeks to analyze how new terms circulating in public discourse on AI become humanized or anthropologized across different linguistic and cultural contexts.

The Department of Digital Humanities at the Université Libre de Bruxelles, with its strong expertise in journalistic corpora, represents a key asset for compiling multilingual datasets drawn from media discourse. These corpora provide a robust empirical basis for examining how anthropomorphic expressions are used, distributed, and contextualized in different languages.

In parallel, computational methods for measuring anthropomorphism have been discussed and refined for cross-linguistic application. Particular attention has been devoted to adapting existing metrics to the structural and grammatical characteristics of French, Dutch, and Italian, ensuring methodological consistency while respecting linguistic diversity.