Scanning the horizon for Singapore’s schools.
An intellectual exercise, run in late 2025 with a few friends who share an interest in where schooling is heading. We scanned the horizon for education in Singapore over a few months, and ended with four scenarios and a set of responses a school or system might rehearse against them. It is written up as method, because the method is the part that transfers.
What this is
A worked example. Everything on the strand page is a claim about how foresight ought to be done. This is one occasion of actually doing it, thoroughly enough that the claims got tested.
Nobody commissioned it, and no decision was waiting on it. A few of us were curious about where schooling in Singapore might be heading, and wanted to give the method a proper run on a question we cared about. In the strand page’s own terms that makes it sensitisation rather than planning, and I have tried to keep the write-up clear about the difference.
I have kept the scenarios in full, including the narratives written to inhabit them. Scenario work is judged by whether the scenarios are any good, and a page about scenario method that will not show its scenarios is asking to be taken on trust.
The shape of the method
We ran it as three interconnected, iterative processes rather than three sequential stages. We collected inputs. We read them against what the education system has said it is aiming for. We generated outputs for application. Then the outputs sent us back to look for inputs we had missed, and the loop ran again.
Reading signals against the system’s stated aims was our own method choice. Nobody asked us to. It gave a global scan somewhere particular to stand, and it has a cost, which the signals page sets out.
Drawn as a straight line, this looks like a pipeline that turns raw material into conclusions. In practice it ran as a loop. The prioritisation step is where nearly all the judgement lives, and an accurate diagram would show most of the arrows pointing backwards.
Why two axes, and why those two
A scenario set is defined by the uncertainties it chooses to be uncertain about. The two axes here were chosen because they are the two forces most likely to determine whether educational technology helps or harms, and because neither is settled.
The first is the state of Singapore society: whether it becomes more cohesive and future-ready, or more fragmented and defensive. The second is how technology is used by the people in it: whether that use is aligned and human-centred, or misaligned and extractive. Crossing them gives four futures that are genuinely different from each other, which is the only test a 2×2 has to pass.
The axes deliberately leave out the rate of technological progress. That was tempting, and it is what most technology scenario sets pick. I think it is the wrong choice for education, because the school-level consequences of a fast timeline and a slow one look far more similar than the consequences of a cohesive society and a fractured one.
Two pages
The work, in two halves
The method and its output. Neither is much use without the other. A scan with no scenarios has produced nothing, and scenarios with no visible method are asking to be believed.
Signal collection
Anyone can collect. The interesting part of a horizon scan is the narrowing: what gets discarded, on what grounds and whether the grounds were stated before or after the discarding happened.
The four scenarios
Two axes crossed, each future written as a day in the life of a student, a parent, a teacher, a vendor and a policy officer.
Output
The four futures
Method
What the set was for
The most useful framing I found came from published foresight work on frontier artificial intelligence by the UK Government Office for Science, which separates three distinct uses. They are worth keeping apart, because an exercise that has not decided which of the three it is doing will drift between them and deliver none.
01
Backcasting
Identify the future you do want, then work backwards from it to a plan that heads towards it. This treats scenarios as destinations, and it is the only one of the three that produces a direction.
02
Stress-testing
Take a response a school or system is already committed to and run it through each scenario in turn. Where does it hold, where does it fail and how would it have to be adapted? Highest hit rate, least glamour.
03
Resilience checking
Test plans and assumptions against shocks nobody anticipated, to see whether they hold up across a range of outcomes or only work for one. This is the use that most often changes a decision, because it surfaces a single point of failure.
Unsettled
What I would still argue about
Specific to this exercise. The arguments about foresight in general sit on the strand page.
- Whether a 2×2 is a method or a habit. Four is the number of scenarios a group can hold at once, which is a fact about people rather than about futures.
- Whether naming a scenario helps or hurts. A memorable name travels. It also hardens.
- Whether it changed anything. No decision hung on it, so the test that matters for planning never applied. The test left is whether any of us read the news differently a year on, and that is the recognition the strand page admits is close to unmeasurable.
- Whether writing the narratives was rigour or theatre. They are the part people remember, and memorability is not the same as validity.