How Generative AI Tools Support the Learning of American Idioms: Insights from Moroccan Postgraduate Students

Abstract

The present study seeks to explore the use of AI tools, attitudes, and perceived idiomatic knowledge among Moroccan students enrolled in EMI programs in higher education institutions.  178 participants, who were Moroccan Master’s students from the department of English studies from Faculties of Arts and Human Sciences, Meknes and Kenitra, participated in the present study. It should be noted that the employed research design involved a convergence approach. In order to accomplish the aim of the research, a well-structured questionnaire consisting of close- ended questions and reflective open-ended questions was applied. According to the results, 86.5% of participants make use of AI tools quite often. In accordance with that, ChatGPT, Claude, and Gemini were mentioned as the most used platforms. As for the attitudinal aspects, all 7 dimensions obtained more than scale average (M > 3.0). Ease of Use (M = 3.72) and Usefulness (M = 3.65) were rated the highest, followed by Retention (M = 3.16) as the lowest dimension. Average knowledge of idiom score was 2.64/5.0, with a sharp reduction in familiarity between idioms with transparent cultural backgrounds ('Under the weather', M = 3.97) and those with obscure cultural backgrounds ('Nailing Jell-O to the wall', M = 1.82). Frequency of use by AI was associated with positive attitude towards AI (rₛ = 0.372, p < .001), but not with idiomatic knowledge; neither was frequency of use by the participants. CEFR level had a trend but insignificant effect on idioms (F(2,175) = 2.889, p = .058). However, there was a significant effect comparing the Intermediate and C2 levels (p = .023). It is noteworthy that even combined, independent variables only accounted for 4.9% variance in idioms scores. Our findings highlight the importance of AI incorporation into EMI education using culturally sensitive approaches such as generating examples and voice interactions.