On the 21st of October, Maria Labarta—a Professor at the Faculty of Philology, Translation and Communication at the Universitat de València—delivered her ninth intervention for the Institute of European Studies, presenting her research to a group of 25 interested listeners. Professor Labarta also has experience reaching undergraduate and postgraduate courses in Spain, Germany, and Brazil, as well as with research in Argentina, New Zealand, Norway, and the U.S.A, having conducted projects both at UC Berkeley and at CUNY.
Discussing “Artificial Intelligence in Foreign Language Education: A Caste Study of German for International Business Students at a Spanish University”, Labarta delved into the nuances of the use of AI in language learning and teaching, attempting to understand student’s observations, attitudes, beliefs, and uses of AI-tools in her introductory-level ‘German for Business I’ class. Labarta focused on four research questions for her study, in which 45 students participated, asking what students report when using Machine Translation (MT) tools, how they perceived the affordances and limitations of such tools, how they reported having learned through classroom-based activities, and how students believed the future of AI MT tools to look. Her research originated from an understanding that students were already using AI tools in class—an understanding substantiated at the outset of the study period, where 90% of the students admitted to already using AI tools to learn languages, citing in particular the speed and ease of AI usage as key factors in their past usage of MT tools.
Through a series of tests including back translations, output comparisons, and group discussions with native German speakers, however, Labarta was able to showcase to the students how AI was inconsistent, and often unreliable in idiomatic contexts. Students additionally remarked on the invisibility of errors generated by AI, as the MT tools don’t admit to making mistakes. Critical engagement with AI, she specified later, is key for future use in language acquisition. Despite this, various benefits to language learning were also uncovered, with students once again referring to the accessibility, speed, and ease of MT services, as well as the variety of possible translations offered by AI systems like DeepL. Additionally, students commended ChatGPT’s explanatory capabilities when they ran into translations they did not understand, or its ability to discern questions of nuance between the differing translations given by MT tools.
Labarta came to the conclusion that, while massively helpful, AI could not yet replace language-learning classes, but that they can co-exist in a symbiotic relationship, given an effective integration of AI tools into the classroom. She also remarked how the experience, particularly given the youth of the participants, allowed her to forge a closer connection to her students, facilitating learning concurrently in that sense. Importantly, she highlighted the significance of explicit instruction of and with AI tools, to create critical engagement and analysis of answers developed by AI, allowing for real, sustained, synergic learning.
Prompted by a question posed by an attendee, inquiring about long-term effects to students learning when MT tools were taken away, Labarta not only affirmed consolidated gains, but also highlighted the effects of MT tools on concentration, energy, and interest in class, citing higher levels of all, during and after the study. Ultimately, Labarta found, and explained, how LLM and MT use in the classroom, when encouraged actively and explicitly, led to lower blind-trust, and fostered critical engagement with AI tools, providing a comprehensive understanding of the possible roles of AI in language pedagogy.