The Trust Imperative - Measured Response #4

Posted by:
Paul Muir, Chief Customer Officer at risr/

Next week the assessment sector decamps to the e-Assessment Conference in London for three days under a banner that is either a statement of confidence or a confession, depending on how you read it: “The Trust Imperative – Innovation with Integrity”.

It is a great theme. It is also a slightly nervous one. You don’t frame an entire conference around trust unless you suspect, somewhat quietly, that you might be running low on it.

And the timing is almost too perfect. In the same stretch of days that we convene to talk about trust, two very different authorities have pronounced on the same anxiety. Ofqual’s chief regulator has warned the watchdog must “move really fast” against a new wave of exam-cheating technology. And the Pope, yes, THE POPE, has used his first encyclical to ask us to remain profoundly human in the age of AI. When the regulator and the Pontiff are worried about the same thing in the same week, it is worth asking what they have both noticed.

So what themes and topics am I most looking forward to exploring next week?

1. The regulator’s arms race

This week Sir Ian Bauckham warned that Ofqual must “move really fast” against a small cottage industry of cheating kit now advertised openly on TikTok and YouTube: AI smart glasses that scroll text across the inside of a lens, earpieces small enough to vanish, even vibrating collars worn around the neck. His underlying point is the one that matters for everyone in the room next week. When help is obtained unfairly, you end up with grades/test scores that no longer mean what they claim to. That is a trust problem dressed as a security problem.

Now hold that next to the other thing Ofqual has been saying: that on-screen assessment for the big-entry subjects probably won’t arrive until the 2030s. There, in a single week, is the sector’s central bind. We must move really fast on the threats and really slowly on the technology. You cannot, it turns out, do both, and pretending otherwise is how trust quietly leaks away.

2. Anticipating change: the 25-year question

Monday’s AI Symposium opens the week, and the session I’m part of alongside Liberty Munson, Leslie Thomas, Ph.D., Marc J. Weinstein and Tim Burnett “Anticipating Change: How AI Will Shape Assessment Over the Next 25 Years” takes the long view across one, five and twenty-five years. The temptation on these panels is to perform either techno-rapture or doom. I plan to resist both, but with panels who knows where we’ll end up. Maybe I should ask AI…

Something I’ve been observing recently, is more and more organisations that treat AI as an accessory and bolt it onto the side of the existing process. Same exam, same item bank, same test centre, now with an added ‘clever’ AI feature. Organisations that treat it as architecture ask the harder question: What would we build if we started from the capability rather than the legacy? That’s the re-innovation test.

And here the Ofqual paradox from above becomes the whole long horizon argument. Traditional testing doesn’t die; it becomes one product among many in a field that finally remembers it’s in the business of competence, not exams. The reason an AI-native competitor hasn’t already eaten our lunch isn’t that the technology can’t, it’s that regulators, professions and incumbents aren’t ready. Or if I was deliberately being controversial, maybe they don’t want us to be ready.

3. The experience gap

RM Assessment opens Day 2 with a keynote on bridging generational expectations in digital assessment, or the “experience gap.”

Here’s why it belongs in a conference about trust. A candidate who finds an assessment confusing, clunky or hostile doesn’t conclude ‘poor UX’. They conclude the test is unfair. Experience is a fairness signal, and fairness is the load-bearing wall of trust. The generation now sitting our assessments has been raised on interfaces that anticipate them. Hand them a 2009 testing experience and you haven’t just annoyed them, you’ve quietly told them you don’t take them seriously. That has consequences for construct validity, not merely satisfaction scores.

I’m really looking forward to that cross-generational opening keynote!

4. From considerations to practice

This one’s also one I’m involved with: Human Oversight of AI in Assessment: From Considerations to Practice again alongside Liberty, Leslie and Marc, but joined by renowned Assessment Technology expert John Kleeman. The sector has spent two years writing thoughtful considerations about human oversight. The question now is whether any of it survives contact with a live operation.

The distinction I most want to land is between oversight that is meaningful and oversight that is ‘theatrical’. It is very easy to put a human in the loop and call it governance. A reviewer who rubber-stamps a thousand AI decisions an hour is not oversight; they are automation bias with a job title. The practice worth having calibrates oversight to risk, human-in-the-loop (or maybe expert-in-the-loop) where the stakes demand it, human-on-the-loop where they don’t, an honest audit trail either way. And the EU AI Act’s Article 14 has stopped being a thought experiment and has become a deadline (albeit extended!). That enforcement ‘clock’ will professionalise our practice faster than any quantity of good intentions.

5. Keep the question of the human in play

Then this week, the Pope. Pope Leo XIV’s first encyclical, Magnifica Humanitas, has I’m sure been appearing within everyone’s LinkedIn feeds, and although it is ostensibly about AI in the round, its through-line lands squarely on us. Worth, he argues, is not earned through efficiency; dignity comes from simply being human, not from output. His closing instruction, “remain profoundly human”, is aimed at all of us, but it could just as easily have been drafted for our assessment audience.

There’s also a small detail I can’t stop thinking about. It took him the best part of a year to write 42,000 words an AI tool would draft before lunch and he’d say that’s precisely the point. For a sector that spends its days measuring human performance, and is now scrambling to certify human authorship, the message to never let the question of the human drop out of view isn’t a homily. It’s a design brief.

6. Guilty until proven human

Which brings me, neatly, to the last of the sessions I’m involved with: Guilty until Proven Human: The Next Conundrum for Test Security, with Isabelle Gonthier, PhD, ICE-CCP and Claire McCauley . We have arrived, almost without noticing, at a model in which the candidate is presumed to have used AI until they can demonstrate otherwise. The burden of proof has been reversed, and nothing corrodes trust faster than telling an honest person to prove their innocence, particularly when the detection tools doing the accusing are confidently wrong often enough that the false positive is now a feature of the system.

And note the uncomfortable ties with the encyclical. The Pope wants us to celebrate the human; our security systems increasingly want us to prove it, under suspicion, with the candidate carrying the burden. Two views of humanity in the same week, one a dignity, the other a defendant. My argument: if your model depends on sorting the human from the AI, you have already lost, because the line is dissolving as we watch.

The Measured Response

The conjunction “with” in the conference theme pairs innovation and integrity as if they require a careful balancing act. This framing suggests two competing values that must be traded off against one another, adjusting a dial until a sensible middle ground is reached.

I think that framing is the trap.

Integrity isn’t the brake on innovation. It’s the reason for it. The whole point of an assessment is that someone can rely on what it says. Strip out the integrity and you haven’t got a leaner product, you’ve got a very efficient way of producing meaningless results. That, in the end, is what the regulator and the Pope both noticed this week from opposite ends of the assessment universe. That the value of the thing rests on something human and reliable, not merely fast. Ofqual frets that the grades won’t mean what they claim. The Pope insists a person is worth more than their output. Same worry, different vestments.

So the measured response to “The Trust Imperative” is to stop treating trust as a tax on progress and start treating it as the specification. Build for it from the first line of the design, not the last line of the compliance review. The organisations that work this out won’t be choosing between innovation and integrity next week. They’ll have noticed those were never two things.

Over to you…

If you’re in London next week, I’d genuinely like to know: what’s the one session you’ve circled in the programme and what would have to happen there for you to leave actually convinced, rather than merely impressed?

The full e-Assessment Association conference agenda is linked here: https://web-eur.cvent.com/event/c667f04c-2eeb-4ebe-9a08-cbf5044422e5/websitePage:5d618159-9336-49d1-9d35-669e340a257a

And if you’re sitting this one out next week: when you hear an assessment described as “trusted,” whose trust are we talking about? The regulator’s, the awarding body’s, or the candidate’s? Because they don’t always want the same thing, and pretending they do is how we ended up with an imperative in the first place.

Come and find me at the panels or the risr/ stand. I’ll probably be the one arguing the burden of proof has ended up on the wrong person.

Comments are open.

Until next week,
Paul Muir

Comment on LinkedIn

The Trust Imperative – Measured Response #4