Understanding the impact of offloading with AI on exam performance and assessment validity

Artificial intelligence (AI) chatbots are transforming education, but their impact on student learning and assessing academic merit remains unclear. This research investigates why students use AI and how using AI while taking online exams affects student performance, whether the outcomes of the exams remain valid, and whether it promotes or hinders learning over the longer term. With the widespread use of AI chatbots by students, understanding the impact of AI on learning and assessment is key to ensuring effective education and evaluation. If successful, results from this project have potential translation to the classroom in helping educators promote learning and design effective online exams in the age of AI. In addition to the research, the project provides workshops for educators and creates research opportunities for undergraduate students in cognitive and computer sciences, preparing them for STEM careers in an AI-driven world. This project frames the use of AI as a form of cognitive offloading and investigates the factors that drive students to use AI chatbots for online exams. Furthermore, the project tests whether or not this leads to gains in performance, and if this promotes or hinders long-term learning. A series of studies are planned that test key predictions that cognitive effort, digital literacy, and content knowledge play essential roles in students’ decisions to use AI tools and also relate to its efficacy. Advanced analyses of student-AI interactions identify what makes some prompts more effective than others for performance and learning. The project also examines whether AI-resistant question designs can help maintain exam validity. The overall goal of this research is to better understand AI in terms of cognitive offloading and how this may be used in educational settings. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.