Stumbling Blocks and Learning Opportunities: Using AI in the Second-Language Classroom

Abstract

This contribution explores three pedagogical stumbling blocks I encountered when employing generative and translation AI in an upper-division German language classroom at Utah State University (GERM 3050: Advanced German Conversation). While the specific context was an in-person advanced German course, many of the challenges—as well as possible pedagogical responses—translate readily to other second-language classrooms and beyond.

The first stumbling block involved students’ reluctance to engage with AI at all. Some of the strongest students in the course resisted using AI tools even in carefully structured classroom activities that invited them to interact with virtual conversation partners calibrated to their respective proficiency levels. Their resistance raised broader questions about trust, authenticity, and the perceived role of AI in language learning.

A second challenge emerged during activities in which students used AI tools to translate short medical dialogues and compare the resulting translations. German is a highly gendered language, and the outputs consistently defaulted to masculine generics and reinforced binary gender stereotypes: doctors became “Herr Doktor,” while nurses were rendered as “Schwester.” None of the translations employed gender-inclusive language. Rather than challenging dominant assumptions, the AI-generated translations reproduced and normalized them.

The third stumbling block arose in connection with AI-assisted translations of museum texts prepared for student activities at the Nora Eccles Harrison Museum of Art (NEHMA). Here, the challenges included inadequate interpretation of imagistic or contextual elements, inconsistent translation choices across related materials, and insufficient adherence to target-language conventions and linguistic coherence, particularly in the use of pronouns and referential structures.

My contribution moves from identifying these challenges—student resistance to AI, gender-related biases in translation, and non-idiomatic or nonstandard language usage—to considering how such problems might themselves become productive sites of learning. Rather than treating AI merely as a tool for efficiency, I argue for a more critical pedagogical approach that encourages students to interrogate AI-generated language, cultural assumptions, and translation practices.

Presenters