Name
Lourdes Hernández Martín ( Language Coordinator, LSE Language Centre)
Email
l.hernandez-martin@lse.ac.uk
Teaching context
LN220 – Spanish Language and Society is proficiency Spanish course forundergraduate students. The 25-26 cohort comprises 10 students from diverse disciplines including Management, Anthropology, International History, International Relations, Social Policy, and Language, Culture and Society. The course combines sociocultural units with a research-based project. This task was delivered in Weeks 4-5 of Autumn Term as part of the sociocultural unit on Power, Identity and Languages in Spanish-speaking contexts.
What was done, and how was AI integrated into teaching/learning/assessment?
I designed a structured task integrating Claude into a text analysis activity on Gloria Anzaldúa's "How to Tame a Wild Tongue." The task alternated between "with machine" and "without machine" activities. Students first set up Claude with a detailed initial prompt establishing context and role. They then used Claude to generate presentations for different audiences, create text summaries, and explore sociolinguistic concepts. Students also had to upload Anzaldúa's text to Claude and compared its analysis with their own reading. The task deliberately exposed students to LLM characteristics such as sycophancy (by having them challenge Claude's responses) and the authoritative voice, prompting reflection on verification practices.
What was the aim of the project?
The task had dual aims: content mastery and academic GenAI literacy. For content, students would analyse Anzaldúa's text and understand concepts like code-switching and translanguaging. For GenAI literacy, I drew on the 4Ds in Dakan and Feller's AI Fluency Framework (for further information see Anthropic's AI Fluency for Educators course): Delegation (D1—understanding when to use LLMs), Description (D2—effective prompting), and Discernment (D3—critical evaluation of outputs). An unexpected third content aim emerged: applying Anzaldúa's critique of linguistic hegemony to Claude's Spanish language production.
What was the projct outcome? Were you satisfied, what did you learn, and what was the biggest challenge?
I am very satisfied with the outcomes. All students engaged with the task, including two who were initially reluctant. Students particularly appreciated the structured approach, the embedded reflections on pros/cons of Claude for academic work, and the connection between Anzaldúa's ideas and Claude's language production. The classroom dynamics worked well despite varying levels of LLM expertise among students. The biggest challenge was the iterative design process - testing prompts myself, consulting with the AI and Education Specialist, Steven Williams, and refining based on classroom experience for future implementation.
What advice would you give to colleagues thinking about integrating AI into their teaching?
- Transparency vis-a-vis students is crucial—explain your pedagogical rationale and acknowledge the experimental nature of integrating LLMs (at this stage, they are new for everyone in education).
- Link GenAI literacy to your discipline's content; this ensures all students find the task relevant regardless of their prior LLM experience.
- Design tasks that require comparison between students' own work and Claude's outputs, as this draws students back into engaging more deeply with source texts.
- Include structured "without machine" activities alongside "with machine" ones.
- Test prompts yourself before class.
- Finally, consider consulting with specialists—the support from the AI and Education Specialist, Steven Williams, was invaluable in the task design.
Documents related to this study
Case study document - Academic LLM literacy through content (pdf)
Task description document for course (Word)