Abstract:

The rapid expansion of Artificial Intelligence (AI) tools is transforming teaching practices across higher education. While these technologies offer new possibilities for learning support, they also challenge traditional forms of evaluation, particularly in disciplines that rely heavily on written interpretation and critical reflection.

This proposal presents a teaching experience from the course Composición Arquitectónica 3 – History & Theory of Modern Architecture (75 students) in the third year of the Architecture Degree Program at the University of Zaragoza (Spain). It examines how the emergence of AI affects written assignments and evaluation strategies. It suggests a necessity for new evaluation models that shift the focus from the final written product to the assessment of the process through oral examination. Instead of adopting a prohibitive or defensive approach with anti-plagiarism software (such as Compilatio or Turnitin), the verification of critical thinking must be redefined, particularly because textual fluency in writing no longer reliably confirms intellectual authorship in the current Generative AI Era.

The course assessment involves two main written components: a series of short theoretical questionnaires (approximately 800 words total per question) and a long-form research paper (3,500 words) on a course-related topic chosen by the student. The evaluation showed a notable gradient in their reliance on AI. This experience combines classroom observation with the results of an in-class survey conducted among students in order to better understand how frequently AI tools are used and for what purposes. While students use AI for initial drafting, its prevalence is notably higher in short assignments than in comprehensive research papers. Our findings indicate that short questionnaires focusing on fundamental historical concepts are highly susceptible to AI substitution, as GenAI excels at synthesizing broad historical data. In contrast, long-form research papers, which require personal argumentation, specific architectural analysis, and close reading of particular bibliographic references, encourage higher individual engagement and are more resistant to “one-click” AI generation. Nevertheless, the use of generative AI should be evaluated to ensure intellectual authorship and to promote the correct use of theoretical terminology and logically coherent critical narratives.

Our model introduces the Oral–Written Feedback Loop as a fundamental tool to engage students in the relevance of critical reflection:
(1) AI-augmented scaffolding: explicitly allowing AI for initial tasks to reduce friction in the early stages of research;
(2) short oral follow-up sessions as a primary verification tool, in which the student must deconstruct and defend the reasoning presented in their writings; and
(3) intellectual traceability, emphasizing the process of architectural analysis—sketches, comparison matrices, and draft iterations—as much as the final polished text.

In Spanish higher education, our experience suggests a future for teaching Architectural History and Theory that moves away from relying solely on the "final paper" as proof of knowledge. We propose an assessment model that integrates written work with a verbalized critical defense. However, this scope necessitates greater time for evaluation, smaller class sizes, and an increased number of teachers participating in the assessment process, which is, as well, a new challenge.

Keywords:

Artificial Intelligence, theoretical, architecture.