The two-lane model, secure versus open, has become the tertiary sector’s shorthand for responding to generative AI in assessment. It is a partial answer at best, and this note is scoped to where it actually applies: written assessment and cognitive learning outcomes, not every kind of task.
Corbin, Dawson and Liu (2025) define securing precisely: a structural change that works whether or not a student complies with an instruction, not a line in the brief telling them not to use an AI tool. That distinction turns the two-lane model into a design decision rather than a policy label, and it comes with a consequence the diagram leaves out: a structurally secured task, Corbin, Dawson and Liu note, often assesses a different learning outcome from the one it replaced. Curtis (2025) adds that security sits on a continuum, not a switch, so most modules need only a small, supervised component.
A contract law vignette works the idea through: one task in a three-part module no longer proved what it claimed to, and fifteen honestly costed, clearly explained supervised minutes did the securing a full exam did not need to.
Reviews
There are no reviews yet.