CLUNL

Raquel Amaro e Tibor Vocásek, investigadores do grupo LLT do CLUNL, apresentam hoje a comunicação “Digital Research Data and Human Sciences: Digital Humanities in the Age of AI” na conferência Digital Research Data and Human Sciences (DRDHum), em Joensuu, Finlândia.

Resumo
The recent buzz around Artificial Intelligence (AI) has raised significant debate in the EU on how to regulate it. Like any emerging technology, AI development depends on mediashaped public perception (Chuan et al., 2019). This research aims to contribute to the study of this perception through an interdisciplinary comparative perspective. It analyses AI regulation representations in the Czech and Portuguese online mainstream media, using Corpus Approaches to Discourse Studies (CADS) (Baker et al., 2008; Partington et al., 2013).
Czechia and Portugal are similar in area size, population, or GDP (Eurostat, 2024) but differ significantly regarding their tech sectors. While news headlines labelled Czechia the “sick man of Europe” due to its stuttering industrial economy (Willoughby, 2023), Portugal, with its vivid tech scene, was referred to as “Europe’s Silicon Valley” (Böhnisch, 2021). Focusing on the EU debate, these countries’ membership differs by almost two decades.
CADS combines corpus linguistics with traditional CDA while reflecting on its critiques (Orpin, 2005). It mitigates issues of data representativeness and interpretative transparency. CADS allows the investigation of the aggregate effects of language, highlighting typical discursive patterns. Conceptually, this study approaches AI regulation through “sociotechnical imaginaries”, understood as “collectively held, institutionally stabilised, and publicly performed visions of desirable futures” (Jasanoff and Kim, 2015, p. 4).
Current studies using imaginaries are plagued by conceptual ambiguity (Richter et al., 2023). This research overcomes the issue by adopting a three-level imaginary concept (Sau, 2021) operationalised with CADS. It structures imaginary analysis by asking for representations of (1) social commentary, (2) vision of the future, and (3) means to achieve it. The analysis covers the period of discussions about the EU’s “AI Act” regulation (3/2018-12/2023).
Comparable corpora, collected from digitally available national media in each country in this period, are compiled and explored using Sketch Engine (Kilgarriff et al., 2014) in two steps. Firstly, distinct imaginary layers are investigated via collocations of the topic-related keywords and analysed via concordances in each corpus. Secondly, the analysis focuses on keywords of the sub-corpus after the ChatGPT onset to reveal specifics of the “ChatGPT-moment”. The results of both steps are then compared against each other.
Although not groundbreaking in terms of methodology, such research provides innova tive, empirically rooted comparative insights into the current media debate on AI. It exhibits the usefulness of the imaginary concept for the CADS while providing a clearer perspective of sociotechnical imaginaries by grounding these to objective linguistic cues.

Saiba mais sobre este evento em: https://sites.uef.fi/drd-hum-2024/