Gerald Ajam

Working propositions

Some ideas are useful before they are finished. A working proposition is a claim that seems coherent enough to test but premature to treat as a principle.

Status Working proposition — Promising, and still being tested.
How I work

Publishing these makes the uncertainty explicit and creates something that can be challenged, refined, contradicted or abandoned.

Colour shows how far testing has got

  • Testing
  • Narrowing
  • Holding

  1. WP-01Testing

    A product becomes educational when its interaction changes what the learner must think about.

    Evidence
    Moderate
    Scope
    Digital learning experiences
    Domains
    • Learning
    • Technology
    • Design

    Current reasoning

    Technology does not become educational simply because educational content passes through it.

    The more consequential question is whether the design changes the learner’s cognitive activity: what must be retrieved, compared, explained, constructed, noticed, decided, or revised.

    What would weaken this proposition?

    Strong evidence that passive interaction structures regularly produce equivalent durable learning and transfer under comparable conditions.

    Related

    Concepts

    • Desirable difficulty
    • Retrieval
    • Agency

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.

  2. WP-02Holding

    Removing friction from an interface and removing difficulty from a learning task are different design operations.

    Evidence
    Moderate
    Scope
    Digital learning products and the tasks inside them
    Domains
    • Design
    • Learning
    • Technology

    Current reasoning

    Friction in an interface is effort spent on the tool: finding the button, recovering from an error, working out what state the system is in. Difficulty in a task is effort spent on the idea. The first is almost always waste. The second is often the point.

    The two are routinely measured with the same instruments, such as time on task and completion rate, so a change that removes the wrong one can look like an improvement.

    What would weaken this proposition?

    Cases where reducing interface friction reliably reduced durable learning, which would mean the two kinds of effort cannot be separated in practice as cleanly as this claims.

    Related

    Concepts

    • Friction
    • Desirable difficulty
    • Cognitive load

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.

  3. WP-03Testing

    The unit of educational design should usually be larger than the digital artefact.

    Evidence
    Thin
    Scope
    Educational technology decisions in schools
    Domains
    • Design
    • Systems
    • Implementation

    Current reasoning

    A learner meets the interface, the task, the room, the assessment and the teacher’s workload at once. Improving one of them while the others pull the other way produces a good artefact and an unchanged lesson.

    The claim is “usually”, not “always”. Some artefacts are self-contained enough to be designed alone. The difficulty is knowing which ones before the evidence comes back.

    Complexity sharpens the claim. The larger unit is not a longer list of parts. It is an arrangement. The same artefact, with the same people, does something different when the relationships around it change, so the thing to design is how the parts act on each other.

    What would weaken this proposition?

    A run of products designed as isolated artefacts that performed as well in varied schools as they did in testing, with no adjustment to the conditions around them.

    Related

    Concepts

    • Implementation
    • Complexity

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.
    2. v0.2, 8 November 2026. Reasoning extended: the unit is an arrangement of relationships, not a bigger set of parts.

  4. WP-04Narrowing

    A scalable intervention must specify what may change as clearly as what must remain invariant.

    Evidence
    Moderate
    Scope
    Interventions moving from pilot to many schools
    Domains
    • Implementation
    • Systems

    Current reasoning

    Every school adapts what it receives. If the design does not say which parts carry the effect, the adaptations fall on whatever is easiest to change, and that is often the part doing the work.

    Naming the permitted variation turns local adaptation from drift into information.

    Complexity sets a limit. In the parts of an intervention whose effect depends on how people respond, the invariant core is itself a guess, because nobody yet knows which parts carry the effect. There the specification should say what is being probed, and what result would move a part from “may change” to “must stay”.

    What would weaken this proposition?

    Evidence that strict fidelity to every component outperforms specified adaptation across varied contexts.

    Related

    Concepts

    • Adaptation
    • Implementation
    • Complexity

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.
    2. v0.2, 8 November 2026. Narrowed: where the effect depends on how people respond, the invariant core is a hypothesis to probe, not a given.

  5. WP-05Holding

    A group needs structures for independent thought before it can benefit fully from collective thought.

    Evidence
    Moderate
    Scope
    Workshops, design reviews and decision meetings
    Domains
    • Collective thinking

    Current reasoning

    The value of a group comes from the differences between its members. Open discussion from the first minute lets the first and loudest view set the frame, and the differences never surface.

    Individual thinking first, written down, protects those differences long enough for the group to use them.

    What would weaken this proposition?

    Consistent findings that groups discussing from the outset reach decisions as good as groups that first work independently, on tasks where members hold different information.

    Related

    Concepts

    • Groupthink
    • Facilitation
    • Participation

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.

  6. WP-06Holding

    The usefulness of an educational metric decreases when the metric becomes a substitute for the construct it was meant to approximate.

    Evidence
    Strong
    Scope
    Measures used to judge learning products and programmes
    Domains
    • Assessment
    • Technology

    Current reasoning

    A metric is chosen because it moves with the thing that matters. Once people are judged on the metric, they find ways to move it that do not move the thing. The number keeps rising and stops meaning anything.

    This is Goodhart’s and Campbell’s point, restated for dashboards. The proposition part is the practical claim: that the decline can be spotted early by checking the metric against a slower, harder measure of the construct.

    Complexity adds that the decline often starts before any stakes are attached. A measure is also a move. Once people can see what is counted, they begin to answer it, so whoever chose the metric is already inside the system it describes.

    What would weaken this proposition?

    Metrics that stayed valid under sustained pressure without any check against the construct.

    Related

    Concepts

    • Assessment validity
    • Learning analytics
    • Evidence

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.
    2. v0.2, 8 November 2026. Reasoning extended: people answer a measure once they can see it, not only once it carries stakes.

  7. WP-07Testing

    The most durable output of an innovation may be the capability the organisation develops while implementing it.

    Evidence
    Thin
    Scope
    Innovation projects in schools and education agencies
    Domains
    • Implementation
    • Systems
    • Futures

    Current reasoning

    Tools are replaced, platforms are retired and pilots end. What stays is what people learned to do while making them work: how to run a trial, read a result, or change a routine without a mandate.

    If that is true, an innovation should be judged partly on the capability it leaves behind, and designed to leave some.

    Complexity names what that capability is worth. It is option value: the ability to change course when the next surprise arrives. Like most option value, it rarely appears in a budget, which makes it easy to spend without noticing.

    What would weaken this proposition?

    Evidence that capability built during one innovation rarely carries to the next, because it stays tied to the specific tool or the people who leave.

    Related

    Concepts

    • Implementation
    • Adaptation

    Revision history

    1. v0.1, 26 September 2026. Initial proposition.
    2. v0.2, 8 November 2026. Reasoning extended: capability as option value, which budgets rarely count.

  8. WP-08Testing

    Most school plans are part complicated and part complex, and they fail when one logic is used to run both.

    Evidence
    Thin
    Scope
    Improvement plans in schools and school systems
    Domains
    • Systems
    • Implementation

    Current reasoning

    A complicated part has many pieces but a knowable answer: a timetable, a procurement, a marking calendar. Expertise and a good specification get it right. A complex part has an answer that only appears through interaction: how students respond to new feedback, how staff use a new rating scale.

    Most plans hold both. Run the complex parts like a timetable and the plan specifies what nobody can know yet. Run the complicated parts like an experiment and nothing gets delivered. The first job is sorting which is which.

    What would weaken this proposition?

    Plans run on a single logic, whether specification or probing, that do as well across their complicated and complex parts as plans that separate the two.

    Related

    Concepts

    • Complexity
    • Implementation
    • Emergence

    Revision history

    1. v0.1, 8 November 2026. Initial proposition, from the complexity strand.

  9. WP-09Testing

    Under uncertainty, a plan should be judged by the options it keeps open as well as by the outcome it targets.

    Evidence
    Mixed
    Scope
    Commitments a school cannot easily undo: contracts, structures and staffing
    Domains
    • Systems
    • Futures

    Current reasoning

    A plan built for one expected future spends flexibility it does not count: the short contract, the spare hour, the second supplier. These look like waste until conditions change. Then they are the only way to change course.

    The claim is not that options should always stay open. Keeping them costs something, and some decisions only work once someone commits. The claim is that the trade should be visible: which doors a plan closes, and which of them cannot be reopened.

    The evidence is mixed in a specific way. The value of keeping options open is well established where uncertainty can be priced. How well that carries into school decisions, where it mostly cannot be, is much less tested.

    What would weaken this proposition?

    Evidence that schools which commit fully and early adapt to surprises as well as schools that keep reversible options, at no greater cost.

    Related

    Concepts

    • Complexity
    • Adaptation

    Revision history

    1. v0.1, 8 November 2026. Initial proposition, from the complexity strand.

  10. WP-10Testing

    A plan’s boundary is a decision, and the plan should say who it leaves out and who bears the cost.

    Evidence
    Thin
    Scope
    Plans, policies and evaluations in education
    Domains
    • Systems
    • Collective thinking

    Current reasoning

    Every plan draws a line around what it counts: which students, which outcomes, which hours of whose time. The line is usually presented as the shape of the problem. It is a choice, and someone outside it pays for it.

    Naming the exclusion does not remove it. It makes the cost something people can argue about, which is the only way anyone can decide it is worth paying.

    What would weaken this proposition?

    Plans that state their exclusions doing no better at avoiding harm to those left out, or at keeping their trust, than plans that do not.

    Related

    Concepts

    • Complexity
    • Participation

    Revision history

    1. v0.1, 8 November 2026. Initial proposition, from the complexity strand.

Singapore