Opinion
The feeling of helplessness that many academic staff report when confronting generative AI (GenAI) is real, but its source is frequently misidentified.
In my view, it is not produced by the technology itself, but by a mismatch between where academic staff believe their control lies and where it actually sits.
Stephen Covey's (2013) circles—control, influence, and concern—offers a useful diagnostic. At the outermost ring sits the circle of concern, the vast territory of things that matter to us but that we cannot act on directly. Inside it sits the circle of influence, where we have no direct control but can shape conditions, shift norms, make arguments, and build coalitions. At the centre sits the circle of control, where we can make decisions and act.
Academic staff who feel paralysed by GenAI tend to be operating from the concern circle. They are worried about what students are doing at home, about whether detection tools work, about institutional policy, about grade inflation, about the downstream credibility of the qualifications their institution awards. These are legitimate concerns. But focussing efforts on things that concern us, without a pathway to action, will, as Covey observed, do "nothing, except to increase our own feelings of inadequacy and helplessness" (2013). On the other hand, academic staff who don't feel paralysed have, often without naming it this way, located themselves in the control circle. They have asked what they can actually decide and act on within their own practice, and they have started there.
An overdue reckoning
What makes the GenAI 'moment' unusual is that it has forced a reckoning with something that was already true long before large language models existed. As I and many others before me have argued, unsupervised written assessment, in isolation, has never provided a defensible evidentiary basis for student learning. GenAI has not created this problem, although it has removed the conditions that allowed institutions and academic staff to avoid confronting it.
For many academic staff, this is the more disorienting realisation, and the anxiety runs deeper than what GenAI systems can produce, touching professional self-concept in ways that go unacknowledged in most institutional responses.
Anxiety, and in many cases, identity panic, arrived with the discovery that their circle of control in assessment was always smaller than they had assumed. When a student submitted an unsupervised written assessment, academic staff believed they were receiving evidence of that student's thinking. The belief was not entirely unreasonable given the available alternatives at the time, but it was never robustly founded. What GenAI has done is remove all "plausible deniability", to quote assessment integrity expert Kane Murdoch. Some academic staff experience this as a loss, and not a minor one. What is lost is not a faulty assessment tool but a professional certainty, and with it a part of the self-image that the assessor role had always supported.
While we’re here, let's put to bed the belief that there's a magical written assessment brief just waiting to be willed into existence by a talented academic that will present a challenge for a GenAI system to complete. GenAI can produce any text, in unsupervised conditions, to a standard indistinguishable from human-written work. Any text. There is no genre, no discipline, no level of sophistication that places a task beyond its reach.
The professional cost
The identity dimension of the disruption rarely gets named directly, and it makes the realisation of lost control considerably more fraught than it might otherwise be.
Academic identity is partly constructed around professional expertise in assessment. The ability to design tasks that genuinely distinguish understanding from its absence, to set the conditions under which learning can be demonstrated and evaluated, is not peripheral to what it means to be an academic. For many academic staff, it is central to it. The assessor's role carries professional authority, and that authority has rested in part on the belief that unsupervised assessment tasks were substantively meaningful as instruments of evaluation.
There is a second dimension to this, one that cuts more personally for those whose academic identity is bound up in the craft of writing itself. Many have spent years, often decades, developing a disciplinary voice, learning to construct arguments with precision and authority, to handle evidence with care, to produce prose that reflects genuine expertise. That craft is not incidental. For many it is the primary medium through which intellectual identity is expressed and recognised by peers. When a large language model produces text that is indistinguishable from that craft, it feels less like a technological novelty and more like a devaluation. The years spent developing something that is now replicable on demand carry a different weight than they did before.
Some of the resistance to accepting the validity problem comes from this identity investment. Conceding that a task can be undetectably completed by AI to a high standard is not an administrative problem to be solved. For some academic staff, it reads as an admission that their professional judgement about what constitutes good assessment was mistaken. The ego cost of that admission is real, and institutions that fail to acknowledge it tend to get compliance rather than genuine engagement with assessment redesign.
The academic staff who move through this most effectively are those who find a way to reattach their professional identity to something more durable than the particular assessment instrument. The question of what genuinely evidences learning, how to create conditions where that evidence is visible and verifiable, how to design assessments that honour both student development and disciplinary standards, these questions have not been answered by GenAI. But they have been made more urgent, and the academic who invests professional energy in them is not diminished by GenAI.
Why some institutional response makes things worse
The institutions that are struggling most with GenAI are those that have treated the whole problem as sitting in the concern circle and have responded accordingly. So-called 'AI detection' tools, tightened academic integrity procedures, process-tracking software, watermarking proposals, and [insert any other technical 'solution']. These responses do not address what assessment experts continue to flag as the key issue: validity. Unsupervised written work, regardless of who produced it, has a limited evidentiary relationship to learning. Knowing who wrote it does not change that relationship.
What these surveillance-led responses do achieve is to move responsibility from institutional assessment design onto individual students and academic staff. Students are positioned as potential violators to be monitored, and educators are asked to police compliance using tools whose outputs are probabilistic and opaque; a role that sits uneasily alongside most academics' sense of who they are and what their role is. The result is an escalation of procedural burden without a corresponding increase in confidence about what student work actually evidences.
I've previously argued that these tools function less as solutions than as stabilising devices, allowing institutions to defer disruptive change while signalling action. But that deferral carries a cost as the longer enforcement substitutes for validity, the harder redesign becomes, and the weaker the assurance that assessment outcomes mean what institutions claim they mean.
Where the agency sits
Academic staff who experience the current GenAI 'moment' as a clarification rather than a loss tend to share a disposition. They have processed, rather than avoided, the professional cost described above. They have accepted that they cannot control what happens outside a supervised environment, and they have stopped treating that as a failure of their professional authority. They have accepted that AI-produced text can rival that of any expert, and they have stopped treating that as a verdict on the value of their expertise. They have moved their energy to the territory where genuine agency sits.
In the control circle, that territory is assessment design at the unit level. The design of what is being asked, how it is structured, and what it treats as evidence of learning are decisions an individual can make and implement without institutional permission. A well-designed supervised task, or an assessment architecture that uses unsupervised work as preparation for a supervised demonstration, does not depend on detecting AI use. It creates conditions where the demonstration of learning happens in a register that is visible and verifiable. The conditions of that demonstration matter as much as the design. Where supervision is available, it shifts the locus of evidence from the submitted artifact to the student's capacity to perform the thinking the artifact claims to represent.
It is worth being honest, however, about what unit-level redesign cannot do on its own. The expert consensus is that valid assurance of learning across a degree requires assessment to function as a coherent program-level system, not a collection of independently redesigned tasks. That kind of change is not within any individual academic's control. It belongs in the influence circle.
In the influence circle, academic staff cannot determine institutional policy and cannot make their colleagues redesign their assessments, but they can model different approaches, make the argument for program-level thinking in the forums available to them, and contribute to the conversations that shape collective practice. The individual who redesigns their unit is doing necessary work. The more consequential move is using that experience to make a visible, evidence-based argument that assessment needs to be understood as a system. Influence is slower and less satisfying than control, but it compounds over time, and it is the mechanism through which individual practice eventually shapes institutional culture.
Redirecting the energy
The helplessness some academic staff feel is real, but it is not a permanent condition. It is the feeling that arises when attention is directed outward toward a ring where agency is absent. And it is compounded by the unsettling discovery that the circle of control was always smaller than it appeared, that the authority over what an unsupervised assessment evidenced was never as robust as the confidence placed in it.
Redesign what you control. Influence what you can. Stop expending energy on the concern circle, because nothing in that ring responds to worry. GenAI has not broken assessment, but it has made visible what was fragile long before anyone had heard of a large language model. That visibility is very uncomfortable, but it is also the necessary precondition for change that actually improves things.
The academic staff who are moving forward are not the ones who have found a way to ignore the concern circle. They are the ones who have accepted its limits and got on with the work that sits inside them. In doing so, they tend to find that the professional identity available on the other side of that acceptance is more durable than the version built on foundations that GenAI has now made impossible to ignore.
The way out is through.
Associate Professor Mark A. Bassett is Academic Lead (Artificial Intelligence) at Charles Sturt University.