Research Foundations

A framework that asks people to seek evidence ought to be able to say what its own evidence is. This page does that, briefly, and is equally specific about where the evidence runs out.

Modern Wizardry is not a scientific theory, a qualification or a claim that every curious person is secretly a polymath. It is an evolving framework with a point of view, and the research below is what that point of view is currently being checked against. Stated as plainly as it can be:

Breadth, depth and integration

The most useful single lens comes from Michael Araki’s doctoral work on polymathy, which treats it as the conjunction of three things rather than as a quantity of knowledge:

• Breadth — genuine engagement with several distinct domains.
• Depth — substantial knowledge or capability in them, rather than passing acquaintance.
• Integration — connecting or synthesising those domains into understanding, action, products or performance.

Each of the three has a counterfeit, and the counterfeits are where the lens earns its keep. Breadth is not a collection of disconnected interests; it is closer to a map than a hoard. Depth is not the same as narrow specialisation, and can be had in more than one place. Integration is not the noticing of an attractive analogy, which costs nothing. And the point of all three is contribution — not erudition performed for an audience.

Two cautions. This is Araki’s model, developed in one body of work; it has not yet been independently replicated and is not a consensus view. And it is used here as a way of thinking, not as a test. His own measure concerns an orientation — whether people value and pursue breadth, depth and integration — and nothing on this site turns that into a threshold, a badge or a definition of who may call themselves a modern wizard. Polymath remains too grand a word for most people to adopt. The research on it is still worth reading.

Connections have to be made properly

Holding several interests does not, on its own, produce synthesis. An experience report by Joan Giner-Miguelez and colleagues, describing two years of social scientists and engineers building research workflows together, is a good account of what it takes instead. Their central difficulty was not technical. It was turning a question asked in one field into something the other could act on.

What helped, in their account and more generally:

• learning enough of each field’s working language to be understood in it;
• translating concepts carefully, rather than assuming a shared word means a shared idea;
• working through co-design, instead of treating one discipline as a service to the other;
• getting ideas out of heads and into models, experiments, workflows or artefacts that everyone can point at;
• testing, documenting and revising what the work concludes.

For a single person working across fields, the equivalent is a short set of questions to put to any connection before trusting it:

1. What evidence supports this connection?
2. Am I representing both fields accurately?
3. What would weaken or disprove it?
4. Does it increase understanding, or merely appear clever?
5. Does it produce useful knowledge, capability or service?

This is the principle of seeking evidence applied to the thing the framework most prizes. Connections other people have not noticed are the interesting part, and for that reason they must stay answerable to evidence — a point made at more length in Seeing Further Does Not Make You Taller. A connection is a hypothesis until it has been tested.

Making is a way of asking

Making is often treated as the demonstration of something learned elsewhere — the model built after the lecture. A small study by Notto Thelle and Lena Vida suggests it is also a way of learning in the first place. They looked at design and engineering courses that use the makerspace at Oslo Metropolitan University, and asked students and their teachers what the work had taught. Three things recurred: bridging theory and practice, collaboration, and learning to use the makerspace itself as a place to learn. Students said that building physical things, and building them again, made them engage with the theory more deeply.

The interesting part is where the difficulty went. Tools that misbehaved, materials that did not do what the drawing assumed, groups that could not agree — none of that was noise around the learning. It was a large part of it. A model that will not stand up has said something the calculation did not. Making, in that sense, is a way of asking reality a question, and reality is under no obligation to give the answer you wanted.

An answer is only useful to somebody listening for it, so the practice here is the whole cycle rather than its exciting middle:

1. Conceive — what you expect to happen, and why.
2. Make — the thing itself, not a description of it.
3. Observe — what actually happened, including the parts you would rather not record.
4. Reflect — why the two differ.
5. Revise — the thing, the idea, or both, and go round again.

Making without reflection is activity. Reflection without contact with reality is speculation. The cycle is the five questions above with the material world doing the testing, and the observing and reflecting are most of what a grimoire is for.

The study is small — ten students and their course coordinators, at one university, describing their own experience. It supports the claim that making can be a way of knowing. It does not show that makerspaces produce understanding on their own: tools and group projects can as easily produce confusion, dependency or a great deal of busy activity when nobody stops to ask what happened.

Orient first. Converse second. Verify throughout.

That sentence is a working rule for learning with AI, and it follows from a study by Hannah Kim, Sergei Kosakovsky Pond and Stephen MacNeil in which a bioinformatics researcher was set a topic well outside their own expertise and learned it with a chatbot as their main tool. The chatbot helped once the learner had a plan and a sense of the territory. Before that, it tended to get in the way.

Orient first means that before leaning on AI in a new subject, it is worth finding out the field’s central questions, its vocabulary, its authoritative sources and its live disagreements. Converse second means using AI to navigate, explain, compare and interrogate — as a well-read conversation partner, not an unquestioned source of truth. Verify throughout means going back to original sources and comparing claims, and it matters most exactly when it is hardest: when you do not yet know enough to recognise a plausible error.

Some friction is worth keeping. Reading the original, finding two authorities in disagreement and working out why — that is not inefficiency to be engineered away. It is a large part of how the learning happens.

The study is small and exploratory: an autoethnography, in which a researcher records and analyses their own experience. It suggests a sensible order of operations. It does not establish a law.

Distributed polymathy

The lone wizard in the tower is a powerful image, and a misleading one. Serious synthesis is very often collaborative. A circle, a workshop, a learned society or a project team can integrate far more knowledge than any one member holds.

Distributed polymathy is this project’s name for that — our interpretation, not a term the research above uses or has tested. The collaboration Giner-Miguelez and colleagues describe is one illustration of it, and their account shows the condition it depends on: it works only when participants learn enough of one another’s languages to think together, rather than simply taking turns to be expert.

It is also why sharing knowledge and enabling other people are part of the practice rather than an afterthought to it. A fellowship, a makerspace, a reading group or a community of people who teach each other is not a support structure for the individual wizard. In many cases it is where the integration actually takes place.

Do It Together

An open door is not the same thing as a welcome. Sonali Hedditch and Dhaval Vyas spent twenty-one months and more than 150 hours of fieldwork in nine Australian makerspaces, looking at what helped refugee and migrant women keep coming back. A do-it-yourself culture, in which everyone is free to enter and expected to find their own way, did not produce inclusion by itself. What did was staff who knew how to teach, learning structured alongside other people, visible routes from a first session to something more, and the confidence of belonging. Their name for it is Do It Together.

That is the practical other half of distributed polymathy. If integration so often happens in a group, the group has to be one people can actually join, and “no gatekeeping” is not enough when the culture still rewards confident insiders who already know the tools and the vocabulary. Whether a circle or a workshop is open is decided by how it is run, not by what it says about itself. That applies with particular force to gender: the wizard here is gender-neutral, and saying so is the easy part.

It is also a correction to a figure this framework is fond of. The self-taught generalist is an appealing image, and celebrating self-direction too loudly hides the mentors, communities, tools and time that made it possible. Independence is a capability worth developing. Isolation is not a requirement.

Across generations

The image this project borrows comes with a default age as well as a default sex, and both are set aside here. Setting age aside need not mean treating it as irrelevant, though. An exploratory report from the UNESCO Institute for Lifelong Learning and Shanghai Open University gathers case studies — from Europe and North America, Africa and Asia-Pacific — of universities and community partners bringing generations together to learn. It treats this as a way of addressing ageism and building age-friendly places, describes benefits to the well-being of older and younger participants alike, and is candid about the barriers such programmes run into.

The risk is sentimentality. An intergenerational programme can easily cast its older members as sages and its younger ones as recipients, which is the bearded figure in the tower again by another route. The version worth having runs both ways. Experience is not infallibility; youth is not intellectual incompleteness; different generations know different things, and teaching can go in either direction. A fellowship can keep knowledge alive without freezing it into tradition, provided it holds to something like this: experience deserves attention, not automatic authority, and novelty deserves consideration, not automatic dismissal.

Curiosity is necessary, and not sufficient

Sustained learning depends on more than temperament. It depends on time, money, confidence, access to tools, good guidance, social welcome, and the absence of unnecessary gatekeeping. Cedefop’s 2026 reference framework for lifelong guidance, written for policymakers across Europe, treats access, support and inclusion as questions of how systems are designed rather than of individual willpower. Its concern is mainly education and careers, and the conclusions drawn here go further than it does.

For Modern Wizardry that means supporting more than one legitimate way of taking part: solitary study; conversation and fellowship; practical making; formal and informal education; fieldwork; artistic practice; experimentation; civic and community service. None of these is the proper route, and none is a lesser one.

Nor does learning have to justify itself through employment or commercial innovation. Wonder, craft, citizenship, fellowship and service are ends in their own right. So there is no single pathway here, and there will be no ranks, grades, badges, compulsory curricula or levels to work through.

Learning is larger than education

The same gap shows up in what gets counted. A 2026 OECD report on measuring lifelong learning argues that existing data systems undercount it: they cover a narrow band of ages, follow few people over time, lean heavily on labour-market outcomes and barely see informal learning at all. It proposes following learning across the whole of a life and across its settings — workplaces, communities and digital environments as well as formal education — organised around the demand for learning, its supply, and the coordination between the two.

That supports something this project has assumed from the start. A notebook kept for years, a program written to see how something works, a dish worked out over a dozen attempts, a season spent watching one stretch of river, a winter of voluntary work: all of it can be learning, and none of it comes with a certificate. It also suggests a distinction worth keeping:

• Practice — what someone actually learns, makes and becomes able to do.
• Recognition — what an institution can see, record or reward.

Modern Wizardry is concerned with the first, which is why it has no levels or ranks. That is not hostility to the second. Qualifications open doors, and a framework that sneered at them would be sneering at the people who worked for them. Recognition is simply somebody else’s job, and done better elsewhere.

The wider definition sets its own trap. If learning can happen anywhere, it becomes tempting to count everything, and repetition, entertainment and passing exposure are not development. So the working rule is this: learning is larger than education, but activity becomes learning only through attention, development, reflection or changed capability. That guards against credentialism, and equally against the opposite mistake of treating every hobby session as evidence of mastery.

Taken together, the sources on this page point the same way. A modern wizard’s practice may be individual, but its opportunities, tools, traditions, challenges and consequences are social. It happens inside a larger ecosystem of learning, and part of the practice is helping to keep that ecosystem open.

What the research does not establish

Precision about limits is part of the argument, so:

• Research on polymathy is still comparatively small, and its central concepts are not settled.
• Much of the work cited here is exploratory or qualitative, and two of the eight sources are preprints.
• The makerspace study rests on ten students at one university, describing their own experience. It supports the idea that making can be a way of knowing. It does not show that makerspaces produce understanding by themselves.
• The Do It Together findings come from one group of women in one country. Applying them to circles and workshops elsewhere is a reasonable inference, not a tested one.
• The intergenerational report is a set of exploratory case studies. It identifies practices and barriers rather than causes, and the insistence on reciprocity above is this project’s position, not the report’s finding.
• Two of the sources, Cedefop’s and the OECD’s, are frameworks for policy and measurement rather than studies of outcomes. They say how systems should be designed and what they should count. Neither shows that any particular informal activity produces learning.
• Research can refine and challenge Modern Wizardry. It cannot prove a cultural archetype, and nothing here should be read as saying it has.
• The wizard framing is an interpretive choice, made for the reasons given on the founding hypothesis. It is not a research finding.
• Claims about Jung’s Wise Old Man, or about archetypes in any psychological sense, would need separate and critical treatment. This page makes none.
• The sources will be reviewed, and the framework revised, as stronger evidence appears — including evidence against it.