Queen Mary's X1-X2-X3 assessment pattern tests AI literacy in database courses

Generative AI has upended traditional assessment, particularly in technical subjects where students can produce plausible answers with minimal effort. In response, Riasat Islam and Thomas Roelleke from Queen Mary University of London developed the X1-X2-X3 assessment pattern, a structured three-part response format that makes AI usage visible and testable. The first part requires students to provide a sourced answer, the second asks them to produce their own response, and the third has them critically evaluate the sourced output — forcing genuine understanding beyond what a simple chat prompt can deliver.

The method also includes an AI-aware question-design process where draft tasks are stress-tested against contemporary GenAI tools and rejected or revised if generic prompting yields superficially adequate answers. Based on a large second-year database systems module, the paper draws on archived assessment materials, rubrics, planning records, design-time GenAI trials, practice-response data, attainment records, and external review comments. The authors emphasize that their main contribution is a reusable assessment-design method, not a claim of measured learning gains. The paper spans 30 pages with 1 figure and 11 tables, and has been submitted to the journal Assessment & Evaluation in Higher Education.