Two-lane approach to assessment

Lane 2 assessments are not simply “unsecured” or “open-book” tasks. They are designed to align with the five action areas in the AI Education Action Plan, supporting students to develop the skills needed in a world where AI is ubiquitous. Lane 2 assessments are authentic, often mirroring real-world or disciplinary contexts where AI tools are already in use. Just as academic staff may use AI tools in teaching, research, and administration, students are expected to use them thoughtfully in Lane 2 assessments.

A better term for assessments that are not fully secured but not yet fully embracing AI might be “towards Lane 2”. In contrast, Lane 1 assessments (assessment of learning) are typically more controlled and may occur in less authentic settings (e.g. exam halls). The rise of generative AI makes securing these assessments more challenging and potentially more artificial.

Not exactly. The distinction is not about the presence or absence of AI, but about assessment conditions and purpose.

• Lane 1 assessments are controlled environments used to verify attainment of learning outcomes. These may or may not involve AI, depending on the discipline and task.
• Lane 2 assessments are more open and formative, supporting students to learn with AI tools where appropriate.

For example, in architecture, generative AI is already used in industry for ideation, so it follows that architecture programs should help students engage with these tools in Lane 2, and therefore aligning learning outcomes to reflect this. However, a Lane 1 assessment might still involve a live, authentic task (e.g., a mock client meeting) without AI support.

Not exactly. Lane 1 speaks to controlled conditions and not to a specific assessment type. There are a broad range of assessment tasks which may be appropriate to Lane 1 conditions, depending on the discipline, purpose of assessment, learning outcomes to be evidenced and the ability to administer, e.g., viva voces, contemporaneous in-class assessments, and skills development, performance and so forth.

Lane 2 provides an opportunity for colleagues to continue to think differently about assessment. We have seen across the University some innovative and creative responses in assessment design. For example authentic assessment that mirrors professional practice, and varied tasks that are adapting to evolving AI capabilities and that require critical and responsible AI use.

Strategic placement of Lane 1 assessments should be coordinated at the faculty level, with input from:

• Programme directors who oversee curriculum coherence
• Associate Deans Learning and Teaching who ensure alignment with faculty assessment policy

This approach supports consistency, reduces duplication, and helps manage workload across programmes.

Lane 2 assessments are primarily assessment for and as learning. The focus is on how students engage with tools, apply disciplinary knowledge, and develop evaluative judgement.

We are not assessing how “good” the AI is, but how well students:

• Select appropriate tools
• Use them effectively
• Critically evaluate outputs

This aligns with University’s Graduate Profile, which emphasises critical thinking, digital capability, and ethical judgement.

Using AI well requires cognitive effort. Students must:

• Understand the task
• Choose the right tools
• Interpret and refine outputs
• Justify their decisions

These are not passive processes. They demand disciplinary knowledge, critical thinking, and ethical awareness, all of which are tested in Lane 1 assessments.

Put simply: if students bypass learning in Lane 2, they will struggle in Lane 1.

 … for example: “AI use is prohibited, or only certain AI tools are permitted or not permitted?”

As currently conceived, no. It is not possible to restrict AI use in assessments that are not undertaken in controlled environments. That is the point of Lane 2; it requires an acceptance that AI may be part of the assessment artefact.

From Liu & Bridgeman, University of Sydney — “It is also not possible to reliably or equitably detect that it has been used, either because a student has access to the latest technology, because they know how to change the raw output, or can afford to pay someone to do this for them. Any unenforceable restriction damages assessment validity so a scale or traffic light approach of telling students that they can only use AI for certain purposes, or use certain AI tools, is untenable. A clearer and more realistic approach is to consider the use of AI in Lane 2 assessments as a menu, where students can pick anything and it is our role as educators to guide them which options are more delectable (better for learning).”

Teachers should consider simplicity/clarity of messaging to avoid complexity/vagary in how this is understood by students.

With respect to academic integrity and AI use, this year the University of Sydney shifted its policy setting to assume the use of AI in ‘open’ or uncontrolled assessments’ — meaning that course coordinators will not be able to prohibit its use in such assessments. Note too that part of the appeal for the two-lane approach is that it has the potential to shift our focus away from policing student misuse of AI, to enabling students to be ethical, discerning and productive users of these tools. It also responds to student concern more widely of being falsely accused of inappropriate AI use. This is not to say that there may be other aspects of academic integrity behaviour and breaches that are still relevant, only that the act of using AI is, in and of itself, not academic misconduct.

In thinking about generative AI, we take the position that it has no agency, so the user (student) who is prompting these tools is to be treated as the author. They are responsible for the work generated by the ‘model’. It is important that authors are aware of the limitations of AI and treat the output critically since they are responsible for that output.

The two-lane approach is built on the principle that using AI appropriately is not cheating.

Detection tools are improving, but they will always lag behind the latest generative models. They are also more likely to catch students who use AI poorly or who lack access to premium tools, raising equity concerns.

Instead of relying on detection, UoA’s approach focuses on:

• Assessment design that integrates AI use transparently
• Clear expectations for students
• Strategic use of Lane 1 assessments to assure learning outcomes