Because responsible proctoring depends on accountable decisions, not automated ones.
Article seven in The Proctoring Question series. The last article argued that access and equity are not the same. This one looks at governance, policy and transparency.
Technology does not make assessment decisions.
People do.
Technology simply enforces those decisions.
Every monitoring rule, every alert threshold, every investigation and every sanction begins with a human decision about what should happen.
A good proctoring platform enables us to verify identity, control access and flag suspicious behaviour, but the credibility of our assessment depends on far more than the features of a platform. Strong governance frameworks place just as much emphasis on privacy, fairness, accessibility and transparency as they do on technical controls.
That is why governance sits at the heart of any credible approach to proctoring. We need to understand who makes decisions, what evidence those decisions rely upon, and how we know they remain justified over time. The answers will ultimately determine whether our candidates’ experience of proctoring is consistent and defensible.
Technology Does Not Govern Itself
I have spoken with several organisations who communicate to their candidates and stakeholders in a way that suggests the technology is making the important decisions. A candidate is flagged, a recording is reviewed, a result is determined, and the process progresses automatically.
The reality is that every one of those steps is shaped by human choices: which behaviours are monitored, what thresholds trigger a flag, which cases are escalated and what sanctions are in place. That human input is not something we should hide. It is something we should be completely transparent about.
The process of governance starts long before a candidate logs into the assessment platform.
We decide which proctoring model is proportionate for a given assessment.
We decide which risks are significant enough to justify monitoring.
We decide what balance we are prepared to strike between security, privacy, fairness and accessibility.
Governance also demands clear ownership. In many proctoring models, responsibility is spread across assessment teams, technology providers, operations, academic decision-makers and compliance functions. Unless accountability for the overall process is explicit, it becomes very easy for everyone to own one part of the system while nobody owns the outcome.
It is governance that turns a set of tools into a defensible assessment system.
Policy Makes Consistency Visible
Policy is where principles become consistent decisions.
It is easy for us to tell our candidates and our stakeholders that we are committed to fairness and integrity. It is much more difficult to show how those commitments translate into rules that are applied consistently across all of our candidates and activities.
A strong policy framework should answer practical questions such as:
- What constitutes a suspicious incident, and what does not?
- How are incidents categorised?
- What evidence threshold is needed before an intervention, investigation or sanction?
- Where AI is used, is it making decisions, or only supporting human review?
- Where, and for how long, are recordings or behavioural data retained, and for what purpose?
- Are remote and in-centre candidates judged against equivalent standards?
From a candidate’s perspective, what matters is consistency and transparency. Without that clarity, even well-intentioned proctoring processes can begin to feel arbitrary.
Transparency Builds Legitimacy
When remote proctoring moved into the mainstream, it was new to us all; we were feeling our way. Today, candidates are much less willing to accept monitoring they do not understand than they were during the emergency shift of the pandemic era. They share experiences, compare processes and have much greater awareness of the questions surrounding privacy, automated decision-making and the use of personal data.
Responsible proctoring demands transparency. We need to be clear about what data we collect, why we collect it, how long we retain it and who has access to it.
Transparency does not mean disclosing every technical detail of a detection model or every internal risk threshold. It means giving candidates enough information to understand the rules of the system they are using.
They have to understand what we are looking for, and the consequences of being flagged.
They have to understand the process by which decisions will be reviewed, and how they can challenge them.
They have to be clear on exactly what sits in the realm of AI, and where the human oversight exists.
When candidates understand the process, they are more likely to accept controls as part of a governed system, rather than as an invisible layer of surveillance operating beyond human control and offering limited scope to challenge. Transparency is what allows candidates to distinguish legitimate oversight from unexplained surveillance.
Governance Does Not Stand Still
No governance model can be credible if it cannot be challenged.
Candidates need clear routes to ask questions, correct misunderstandings and appeal decisions that affect their progression or reputation. We also need robust, regular internal review mechanisms to spot patterns: repeated false positives, inconsistent incident classification, emerging forms of malpractice, or evidence that particular groups are being disproportionately affected.
This is where governance becomes a living process rather than a static policy document. Good governance is not simply receiving a report, noting its findings and moving on. It requires us to ask whether the controls are still working, whether the evidence still supports them, and whether the assumptions behind them remain valid.
Without that review cycle, policies quickly become outdated. A model that was proportionate two years ago may no longer be justified if technologies, risks or candidate expectations have changed.
Good governance is not just about having rules; it is about proving that those rules remain proportionate and evidence-based over time.
Human Oversight in an AI World
The governance question is no longer simply whether AI should play a role in proctoring. It already does. The more important question is where its authority should end.
The more powerful the technology becomes, the more important those boundaries are. Automation may increase efficiency, but it does not transfer accountability from the organisation to the system.
AI may help us to identify patterns, prioritise reviews or detect anomalies, but where a decision can directly influence candidate outcomes, there must be clarity about whether a human reviewed the evidence, what standard was applied, and how the candidate can contest the decision.
In a world where candidates, stakeholders and regulators can all be sceptical about the involvement of AI in high-stakes assessment, it is more important than ever that we can demonstrate that our systems do not discriminate, and that access, privacy and fairness are considered from the outset as fundamental parts of the design.
A black-box alert that cannot be explained, reviewed or challenged does not strengthen or simplify our governance model. It weakens it, generates suspicion and exposes the organisation to greater risk.
A Final Thought
Good governance is not glamorous.
It is not something we talk about at graduation ceremonies, and it rarely appears in product demonstrations or vendor feature lists. But it is what allows us to prove that what we have in place is not simply a system, but a defensible process.
Technology can monitor.
Technology can record.
Technology can flag.
Only people can decide what those things mean.
Good governance ensures those decisions are proportionate, consistent and accountable.
That, not the technology, is what earns trust.
More in this series
The Proctoring Question began with Designing Trust: The Next Chapter of Remote Proctoring, which set out the eight themes that decide whether remote proctoring earns and keeps trust.
Previously: Access and Equity Are Not the Same on why equity begins after login, and the design choices that make it possible.
Next: AI, human judgement and the future of trust, and how we can blend human and digital capabilities fairly, ethically and responsibly.
If you are rethinking how proctoring decisions are made and reviewed in your own organisation, speak to one of our experts. We work with assessment organisations on exactly these questions, and an outside view can help you see where your policies are clear, where they are inconsistent, and where they may be hard to defend.
