by Dana Hanna, Associate Dean (Education)*
Artificial Intelligence (AI) represents a transformational shift to the way we work, learn and produce knowledge. Education has experienced many such shifts over time: the calculator, personal computers, and widespread adoption of the internet changed how and what we teach.
AI is doing the same.
While a challenge, it’s not one that will make the role of an educator redundant, but rather will change what we focus on and how students demonstrate learning.
This post springs from a session delivered by Ryan Payne and covers:
Evolving the way we teach and assess
At a recent presentation to CBE colleagues (AI in Education: Innovating & Transforming Practice – CBE Education), Ryan Payne explained how large language models work using a simple analogy. We were each given identical Lego sets and asked to build a duck. The pieces were the same for everyone (the training data), and construction rules were the same (the algorithm), yet the final ducks all looked different!
AI works in a similar way: the underlying data and models may be shared, but the outputs are shaped by the prompts, the context and the user.
Importantly, AI tools also adapt to the user. Over time they can begin to produce responses they think we want – raising the possibility of confirmation bias and reinforcing existing assumptions.
This has important implications for teaching and assessment.
For me, this emphasises the point of being explicit with our students in terms of usefulness of AI and discernment. As academics we can (often) quickly discern in our own fields information that is correct (or not), but as students, still in the development stage, this is much harder. As such, it’s important to help students navigate AI and develop disciplinary judgment.
Teaching students now also involves teaching (and modelling) appropriate AI use by:
- Verifying claims and references
- Triangulating sources
- Recognising hallucinated information
- Understanding the limits of AI-generated analysis (and bias)
- Using AI ethically and transparently.
This is about explicitly emphasising testing, fact checking and confirmation of resources, i.e. interrogating AI output.
It’s also worth acknowledging that remembering concepts, facts and figures is useful. Yes, you can look it up, but could be embarrassing when your future boss asks you about something and you can’t answer without pulling out your phone and checking…
We are past the point of “AI-Proofing” assessment
There is increasing recognition that attempting to AI-proof assessment is futile, or at the very least doesn’t follow the marginal benefit = marginal cost principle (putting my economist hat on). A more useful framework to think about assessment is using the two broad categories as defined by experts from USyd.
Lane 1 assessments provide assurance of learning.
These typically involve invigilated or supervised formats such as:
- Exams
- Oral assessments (viva voce)
- In-class problem solving
- Interactive orals
These tasks are designed to verify individual understanding of course learning outcomes
Lane 2 assessments assume that AI and other digital tools are available (and being used).
These tasks focus on learning, practice and application. Students develop skills, experiment with ideas and engage in real-world problems. Developing a growth mindset in these activities can prove useful and important.
AI can/should be integrated as part of the process – much as we would use calculators, computers and software applications in the professional environment.
It is these tasks that might form the majority of assessment design in a course (by task number).
Shifting the mindset: Learning first, assurance second
For many of us, academics and students alike, this requires a mind shift.
Rather than designing every task to prove learning, we can think of assessment as having two purposes:
- Lane 1 – assessment OF learning
- Lane 2 – assessment FOR learning
Lane 2 tasks allow students to practice, experiment and build competence (and confidence). Lane 1 tasks confirm whether that learning has actually occurred.
If Lane 2 takes are well designed, students should approach Lane 1 tasks with confidence.
The challenge is alignment: the skills practiced in Lane 2 must clearly prepare students for Lane 1.
What does effective Lane 2 Assessment look like?
Lane 2 assessment often expects higher order of thinking for our students – emphasising process, reasoning and application (as compared to just the right answer).
Examples include:
- Process-based explanations
Students explain how the reached a conclusion, including the assumptions, reasoning and alternative approaches considered. This can do hand in hand with why this is the chosen solution.
- Applied Case Analysis
Using theories and frameworks from class, students analyse real-world examples and evaluate the relative strengths of different approaches.
- Debates and structured discussions
Students defend or critique policy, business decisions or points of view using evidence and theoretical frameworks from class. This can be done in person, in tutorials or using the Canvas Discussion tool for an asynchronous option.
- Role playing and simulations
Students take on roles (consultant, regulator, manager) and respond to unfolder seminars. Again, this can be done in class, but also online or asynchronous).
- Interviews with a summary/blog post
Students interview people of interest, submitting the video, but also a reflective essay or blog post summarising key insights.
- Artifact creation
Students produce tangible outputs demonstrating their understanding. These can include full business plan, art piece, website, infographics
- Competency based assessment
Students attempt assessment which allow multiple attempts at achieving a pre-determined level of competency. These can include online quizzes, tutorial quizzes.
Capturing the Learning Process
With the use of AI tools, the learning process becomes particularly important.
This can be achieved through:
- Draft submissions or staged assessment
- Annotated AI prompts
- Short reflections on how AI was used
- Version histories or research logs
- Peer review activities
Designing for authentic practice
In talking with Industry stakeholder from business, economics and statistics disciplines, we know AI tools are already embedding in professional workflows.
Assessment can therefore aim to mirror workplace tasks, such as:
- Consulting reports
- Policy briefs
- Market analyses
- Board presentations
- Decision memos
Students might even be asked to evaluate AI generated outputs – identifying weaknesses, biases or missing information.
In doing these types of tasks our students can develop the judgment required to work effectively alongside AI systems.
Modelling responsible AI use
All of this means that we too need to be more familiar with the tools.
Understanding how AI systems work – and where they fail – allows us to set clear expectations around AI use for our students, model it’s use (openly) with our students, and design assessment that meaningfully incorporates AI.
AI training
ANU staff have the ability to complete a Certificate in Higher Education. The certificate provides practical insights to integrate AI ethically and effectively into your work. Participants will gain a foundational understanding of AI, explore real-world applications in higher education, consider ethical and inclusive practices, and learn how to manage change and stakeholder engagement.
The program takes around 20 hours and includes core modules plus specialised electives. Completing the certificate helps staff support AI integration, governance and student success across ANU. It will be available until the end of this year (2026).
A number of resources from the Digital Education Council can be found here: Digital Education Council | Education & Training.
*And yes, AI helped me write this. After I had a first draft I asked AI if I was missing anything (it added a few of the workplace tasks) and to put it in a blog format. It also added a Final Thought section. I reworked it again to my satisfaction. Did it save me time? Maybe not. But given I’m new to writing blogs, it did help with the formatting into lots of smaller paragraphs!