The coordinates that reorient learning in an AI-enabled world

Guugu Yimithirr, an Aboriginal language of Far North Queensland, Australia, uses absolute directions rather than body-centred relative terms, such as front, back, left or right. Instead, speakers use directions like north, south, east and west to describe spatial relationships, even when indoors or with no clear landmarks. Over time, their attention to and organisation of spatial information changes – moving from self-oriented to world-oriented frames of reference.

This linguistic phenomenon offers an important insight for education: it is not simply about imparting content but about continually shaping how learners understand the world, orient themselves within it and make judgements in complex situations. 

• Move beyond technical skills and embed AI literacy that teaches students to question
• ‘The more I improved my instructions, the better they worked as AI prompts’
• If a chatbot can pass your assignment, you are grading the wrong thing

In the age of AI, cultivating this capacity for orientation and judgement becomes particularly urgent. Standard answers lose some of their authority when information is so readily available, shifting the role of educator. We are no longer just transmitters of knowledge, providers of correct answers or implementers of curricula. Instead, we must act as architects of cognition, guides for judgement and designers of learners’ cognitive experiences.

Continuously orienting

My university’s adoption of an attainment model made these shifts concrete. As more than 50,000 study plans were created, over 2,000 courses were redesigned and 33,000 timetabling activities were scheduled for the new institution, educators had to reconsider not only what individual courses teach, but how learning is sequenced, judged and connected across programmes.

The reform raised a practical question for educators across disciplines: how do we define, recognise and support meaningful learning when knowledge, pathways and expectations are being realigned at scale?

Revising content was often more straightforward; the more complex task was agreeing what students should progressively become capable of judging and doing. We needed shared ways of understanding progression, judgement and how learning connects across programmes.

Orienting learning beyond the self

Educators need to move beyond self-oriented approaches to educational practice, organised around a single course, a single authority or a single moment of evaluation. Instead, educational design should orient learning toward broader structures of programme coherence, shared responsibility and learning progression. Here’s how to do it.

Beyond individual courses

Begin course planning by explicitly mapping how the course contributes to the broader programme or learning pathway, rather than focusing solely on content coverage. Make this visible to students so they can understand how this course connects to what they have learned before and what they are expected to develop next. 

One example comes from “Generative artificial intelligence”, a course I have taught and continually redeveloped since the beginning of 2026. It plays different roles across three postgraduate programmes, which means course design cannot begin with content coverage alone. 

I also need to consider what students are likely to know when they arrive, what capabilities they need to develop here and what later courses will expect from them. So, the seminar materials are organised to connect prior learning with what follows, rather than treating the course as a self-contained body of knowledge.

Beyond individual activities 

Design learning activities that distribute intellectual authority across the classroom. This shifts learning away from a single authoritative voice and encourages students and the whole teaching team to take responsibility for constructing and evaluating knowledge together. 

In my own teaching, this means treating both students’ responses and tutors’ judgement as part of the learning process. When I teach the same course across different campuses, I use the questions and misunderstandings from the first delivery to adjust explanations, structure and, sometimes, the teaching materials before the next. 

Students are therefore not simply recipients of a finished course, and tutors are not simply delivering content; both contribute evidence that reveals where my assumptions do not match learner understanding and helps me continually recalibrate the course. 

Beyond one-off evaluation

Design assessments that progressively require greater judgement, independence and perspective-taking, while making explicit how each task builds on earlier learning and prepares students for later study. Rather than treating assessment as a series of isolated end points, design tasks that help learners monitor their progress and take increasing responsibility for their own judgement. 

In my generative artificial intelligence course, this progression is built into the assessment structure. The course is organised into three modules: foundations of generative AI, generative models and applications of generative AI, with each stage building on the previous one. Each concludes with a secure module quiz, but the three quizzes together account for only 15 per cent of the final grade. They act as checkpoints, giving students and the teaching team evidence about whether the foundations needed for the next stage are in place. 

Secure assessment is important but relying too heavily on invigilated, time-limited assessments can reduce variety and authenticity. Each module also connects to a more substantial assignment designed to support authentic learning, collaboration, appropriate AI use and professional practice. Clear criteria and structured submission templates are deliberately designed as part of the learning process, helping students judge their own progress while they work independently.

Anchoring judgement in a values-based north star 

Move beyond assessing answers to assessing the processes of thinking and judgement. When we reduce evaluation to measurement, it operates on the assumption that conclusions are already settled and correct answers can be specified in advance. Under this logic, evaluation becomes a technical exercise, applying objective indicators to record performance. 

Assessing judgement, in contrast, requires a different set of questions: how well, against what and why. The value of judgement lies in its capacity to be shared, discussed and explained. 

Treat evaluation as interpretation, not measurement alone. Use quantitative indicators as sources of evidence, rather than treating them as complete evaluations of quality. Scores or completion rates can describe aspects of performance but they cannot by themselves determine the meaning, significance or educational value.

For group assessment in my generative artificial intelligence course, a shared submission does not automatically produce a shared mark, because the final product alone cannot fully show what each student understands or has contributed. In the viva, each student presents their contribution, answers questions and defends the decisions made. The submitted work therefore becomes one source of evidence, while the viva helps interpret that evidence in relation to individual understanding and judgement. Students in the same group may receive different marks according to their contribution and demonstrated understanding. The emphasis shifts from measuring the quality of a single group product to interpreting how each student reasoned, contributed and justified their decisions.

Ensure that every judgement is evidence-based, context-sensitive and justifiable. Consider whether the standards themselves are appropriate, whether the evidence is sufficient and whether the resulting judgement can be justified. Human judgement, grounded in the core values of higher education, should not be reduced entirely to automated responses. 

Ground judgement in collective programme-level expectations, disciplinary norms and professional or ethical criteria – and make these foundations explicit. 

In the final module of the generative artificial intelligence course, which looks at how generative AI is used in practice, I ask students to think more carefully about where human judgement still matters. I walk students through cases showing how a range of AI tools can produce convincing but unreliable or biased answers, and how their outputs can be manipulated through prompting. 

Rather than treating these only as technical problems, I encourage students to consider what evidence would be sufficient before trusting an answer, what risks matter in a particular context and what professional or ethical standards should guide the decision to accept, question or reject an AI-generated response. 

The point is to help students see that using AI well is not just about producing a better answer but about making decisions that can be justified against shared disciplinary and professional expectations. As AI systems become more capable, the ability to judge the quality, appropriateness and consequences of their outputs against these standards becomes even more important.

Education as the art of discernment 

Education is often described as both a science and an art. In the age of AI, one of its most distinctive contributions lies in discernment – the ability to sense meaning, value and depth beyond what can be easily generated. AI has made content flourish as never before, flooding learning environments with information. It’s up to us to help learners develop the capacity for judgement and locate themselves within shifting knowledge landscapes.

Lin Yue is lecturer at the University of Adelaide.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.