Background
Generative artificial intelligence has challenged the capacity of traditional assessment to provide reliable evidence of individual student learning. In particular, unsupervised take-home tasks may increasingly assess a combination of student knowledge, digital literacy and access to external tools rather than the intended learning outcomes alone. This creates a need to reconsider not only how academic integrity is protected, but how assessment is designed.
Methods
Several assessment changes were implemented across biomedical science courses with the aim of increasing authenticity and confidence that submitted work reflected students’ own capability. These included moving selected take-home activities into supervised classes, introducing examinations delivered through Cadmus using a lockdown browser, and replacing a conventional take-home laboratory report with an in-class laboratory simulation followed by report writing under supervised conditions. A student-partnered advisory group was also used to evaluate proposed assessment approaches and provide feedback on their perceived fairness, feasibility and educational value.
Results
These experiences suggest that technological restriction alone may have limited capacity to address the challenges created by generative AI. Lockdown environments can reduce access to external resources, although these may also introduce complexity without necessarily improving the authenticity of the assessment task itself. In contrast, supervised activities requiring students to interpret information, solve problems and communicate their reasoning in real time appeared to provide more direct evidence of individual learning.
Conclusion
Assessment reform in the age of AI may benefit from moving emphasis from detecting inappropriate tool use to designing tasks in which student capability is directly observable. In some contexts, increasingly sophisticated technology may not be the solution, and even pen and paper may provide a more authentic measure of learning.