Restructuring the essay in a time of AI

12pm-1pm (AEDT) | Tuesday 3 March 2026

PRESENTATION ABSTRACT

Generative AI has obviously destabilised assessment. Perhaps nowhere else is this more visible than in the context of the essay. This Essential Insights session focuses on the essay in a time of AI and argues that AI and assessment constitute a wicked problem. This means that there are no ‘correct solutions, only better or worse options, and it also means that technical fixes intended to restore the essay through means such as AI detection, policy, or surface redesign are likely to be ineffective and even counterproductive. Using the essay as a focal case, the discussion shows how GenAI exposes deeper questions about what essays are meant to assess and why we should resist abandoning them. The session concludes by moving beyond diagnosis to outline concrete structural changes to essay tasks that might preserve both their educational value and assessment validity in the age of AI.

Dr Thomas Corbin, Research Fellow at Deakin University’s Centre for Research in Assessment and Digital Learning (CRADLE) will lead this session as a guest speaker.

This session is facilitated by DASSH’s Teaching and Learning Network led by conveners Dr Steven Murdoch, Professor Amanda Mooney and Dr Tony Fisher.

Highlights and key takeaways

Generative AI is changing how students write, how educators assess, and how universities think about academic integrity. Many current approaches are built around a simple idea: drawing a clear line between “student work” and “technology use.” That boundary is getting harder to maintain.

AI tools are becoming more capable, common, and integrated into everyday life. This includes emerging technologies like AI-enabled wearables which make it increasingly difficult to tell when and how tools are being used. As a result, approaches that rely heavily on detection or strict containment are unlikely to remain effective for long.

Instead of treating AI as a problem with one clear solution, it can be more useful to see it as a wicked problem: complex, fast-changing, and shaped by how we define it. If we frame AI primarily as a cheating issue, the response will focus on policing and enforcement. If we frame it as a learning and assessment issue, the response shifts toward redesigning tasks so they remain meaningful in an AI-rich world.

Why essays are under pressure

Traditional essays are facing a legitimacy challenge because AI can now produce polished, convincing essay-style writing. This raises a key question: if an AI system can produce something that looks like a strong essay, what exactly does an essay still prove about a student’s learning?

At the same time, essays can still offer distinctive educational value. Essays can support:

Exploratory and reflective thinking
Self-regulated learning
“Thinking at the border,” where ideas are tested, revised, and developed over time

The challenge is that many modern essay assessments focus mostly on the final product rather than the learning process that leads to it.

Moving forward: designing assessment for an AI world

Rather than banning AI outright, or relying on tools to detect it, a more durable strategy is to redesign assessment so that it remains valid and educationally valuable even when AI is present.

Promising directions include:

Task design that invites distinctive responses
Create prompts that encourage divergence, originality, and disciplinary judgement, rather than highly predictable formats.
Process visibility
Make learning “on the way” assessable by building in drafts, checkpoints, reflections, or other ways students show how their work developed.
Exemplar-based learning and dialogue
Use strong exemplars and conversation to clarify standards, instead of placing all weight on rubric interpretation.
Holistic judgement
Support assessors to make expert, disciplinary judgements rather than reducing evaluation to checklist compliance.
Feedback that recognises developing thinkers
Position feedback as part of learning and identity formation, not only correction.

A realistic goal: clarity without pretending certainty

In practice, there will not be one universal rule that defines “appropriate AI use” across all tasks, disciplines, and contexts. Students and educators often experience a growing interpretive burden as they try to guess what is acceptable.

A productive goal is to build shared expectations locally, reduce guesswork through clear examples, and design assessments that do not depend on an increasingly fragile boundary between people and technology.