2026 ASEE Annual Conference & Exposition

While tools such as ChatGPT and Copilot have introduced new opportunities for teaching and learning in higher education, they have also become shortcuts for students seeking to bypass homework and take-home exams. A small but noticeable minority of students bypass assignments by copying questions into a chatbot and submitting the outputs with little or no learning or critical thinking. Such low-effort use of generative AI tools raises legitimate concerns about the rigor and validity of existing assessment designs. AI-generated answers often include details irrelevant to course content, non-functional formulas and code, incorrect variable names and references, or incorrect answers nearly matching ChatGPT’s basic output. Such patterns are often distinctive enough for instructors to recognize, as the answers contain technical errors and inconsistencies rarely seen in authentic student work. To deter and detect such misuse, the development of the Strategies and Practices to Overcome Irresponsible and Low-Effort Deliverables (SPOILED, pun completely intended) framework was initiated. In its current stage, the framework assesses the vulnerability of Excel- and SQL-based assessments to low-effort misuse of AI tools. The framework also examines how much contextual data is needed for ChatGPT to produce acceptable answers. Initial findings indicate that Excel formula writing questions are vulnerable when students provide complete data tables, but can become resistant with small design choices such as shifting table addresses. SQL query writing questions are vulnerable when students provide ChatGPT with the database schema diagram. These findings support the use of course-specific formatting requirements as well as proctored quizzes and exams to verify student proficiency in Excel and SQL. Beyond these early results, the SPOILED framework provides instructors with a consistent approach to deter low-effort AI usage across different courses. Future work will extend the framework to question types used in a wider range of engineering education assessments. The study underscores the importance of systematic research on responsible AI use and the development of evidence-based methods to mitigate low-effort AI misuse in higher education.