Curriculum change, AI and the race for 'Guild Knowledge'

Across most established curricula, something important already exists but is rarely named. Over time, teachers, examiners, parents and students have built a shared intuition for what “good” looks like: what a strong essay feels like, what a secure piece of maths work looks like, what “ready for the next stage” means in a subject. That shared intuition is what we have been calling Guild Knowledge.

It is not perfect and it is not evenly distributed, but it has been built, slowly, through repeated exposure to real work, informal moderation in staff rooms, past papers, examiners’ reports, and years of lived experience.

Now we are asking everyone to move. Fast.

Curricula are being redesigned to reflect new competencies and higher‑order skills, as projects like the OECD Future of Education and Skills 2030/2040 make clear. Systems are trying to respond to curriculum overload, to close the gap between curriculum intent and classroom reality, and to align teaching, assessment and new societal demands. At the same time, AI is rapidly changing what counts as worthwhile work and what kinds of tasks are even assessable by humans alone.

We are effectively asking students, teachers, parents, employers and everyone else to quickly grasp what quality looks like in a post‑AI world, often while the ink on the new curriculum is barely dry.

That is not just a policy challenge. It is a Guild Knowledge challenge.

The uncomfortable truth about curriculum change

Curriculum documents can move quickly. Guild Knowledge does not.

When a government publishes a new framework, it typically does three things: it sets out new goals, it describes broad pathways towards them, and it signals what should matter more and what should matter less. But it does not, and cannot, fully answer the day‑to‑day question that teachers, examiners and students immediately face:

“What does good work under this new curriculum actually look like, in my subject, with my students, this year?”

In practice, people answer that question by leaning on their existing Guild Knowledge. If nothing actively shifts that internal sense of quality, three things tend to happen, which the OECD’s work on curriculum redesign has documented repeatedly:

First, the written curriculum moves faster than the taught and assessed curriculum. New language appears in policy, but classroom tasks and grading standards quietly stay close to old expectations, because those are what people know how to recognise.

Second, implementation fragments. Different schools, regions and examiners translate the same document into different mental models. For some, a new “competence” means re‑labelling existing activities; for others it triggers genuinely new work. Students’ experience of the “same” reform diverges, with equity implications the OECD has been warning about under its Education 2030 strand.

Third, trust erodes. Parents and employers struggle to read new reports and certificates. Teachers feel they are working under shifting expectations. Young people are left trying to decode what new credentials actually signal in terms of capability.

AI intensifies all of this. As tools become competent at routine writing, calculation and recall, the value of human work shifts towards judgement, synthesis, originality and collaboration. But unless systems deliberately build new Guild Knowledge around those things, assessment will default to what is easiest to mark rather than what is most worth learning.

Why Guild Knowledge has to be built, not assumed

For years, curriculum policy has carried an implicit assumption:

If we articulate the new curriculum clearly enough, people will work out what good looks like.

The OECD’s international curriculum analysis suggests that this is optimistic. Countries report the same issues: curriculum overload, long time lags between reform and impact, and persistent gaps between intended and implemented curricula. Behind that pattern lies a simple fact: documents and rubrics alone do not create a shared internal sense of standard.

Guild Knowledge is formed through exposure and practice: seeing many examples of work, making judgements, hearing what others think, and adjusting over time. If we want a new curriculum to exist as more than words on paper, we have to design experiences that deliberately build new Guild Knowledge for teachers, examiners and students.

The encouraging news is that research shows this can be done.

What we know from Learning by Evaluating

Work on Adaptive Comparative Judgement (ACJ) and “Learning by Evaluating” has shown that people can develop more accurate, aligned judgements through relatively short, structured experiences where they compare real work and decide which is better.

In one large‑scale study with 550 university students on a design‑thinking course, Bartholomew and Mentzer gave half the cohort a 20‑minute ACJ activity before they did their own assignment. Those students spent their time viewing pairs of anonymised peer statements side by side, choosing which was stronger each time. Afterwards, when everyone produced their own work, the group who had gone through this “Learning by Evaluating” activity performed significantly better than the control group, across the ability range.

Other studies synthesised in systematic reviews of ACJ research show similar patterns in schools: when learners, including younger and lower‑achieving students, repeatedly judge authentic work against a standard, they become more reliable judges and their own work improves. The important detail is that the intervention is not more explanation of criteria. It is guided practice in judging real work.

For curriculum reform, that matters. It suggests that Guild Knowledge about new objectives does not have to take a generation to emerge. It can be accelerated, if systems can give large numbers of people structured, feedback‑rich opportunities to evaluate work under the new intent.

Our response: RM Compare and RM Compare | ⏱️NOW

RM Compare exists to turn that insight into something systems can actually use when they redesign curricula.

At one level, RM Compare is infrastructure. Using Adaptive Comparative Judgement, expert groups can compare hundreds or thousands of pieces of student work to build statistically reliable scales (“rulers”) for new constructs – what a particular competency looks like at different levels, what progression looks like in genuine examples. Those scales can then be governed and shared across a system, so that assessment reforms are grounded in a common understanding of quality, not just in text.

But we have learned that infrastructure is not enough. If curriculum change is a Guild Knowledge challenge, we need a way for individuals – teachers, examiners, trainees, students – to feel their own judgement shifting in relation to new standards.

That is why we built ⏱️RM Compare | NOW.

⏱️NOW is a deliberately small tool with a big job. It gives any practitioner or learner a short, on‑demand experience in which they:

• Make an initial judgement about a piece of work.
• Compare it against other examples drawn from a ruler for a new curriculum objective.
• See where it actually sits, and how close their own estimate was.

Nothing is stored; no one is “assessed” in the formal sense. The point is that each loop tightens the user’s internal sense of the new standard. Used repeatedly, and combined with the standards built and governed through the wider RM Compare ecosystem, it becomes a practical way to move Guild Knowledge in step with curriculum policy, rather than years behind it.

Why this matters now

Most governments today share similar goals: to design curricula that prepare young people for uncertain futures, to make curriculum implementation more evidence‑based and systematic, and to align teaching, assessment and new demands from society. The OECD’s Education 2030 work has helped clarify the “what” and the “why”.

Our argument is about the “how”.

If we accept that Guild Knowledge is the hidden engine of any curriculum – the way people actually recognise quality day to day – then we have to treat it as something that can be designed for, developed and governed. That is what we are trying to do with RM Compare and RM Compare | ⏱️NOW: provide tools that make Guild Knowledge visible enough to work with, and flexible enough to evolve as curriculum and AI reshape what matters.

Curriculum change will always be difficult. But it will be impossible if we continue to assume that everyone will simply work out “what good looks like” for themselves, in a world that is changing faster than ever.