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.
01
Collection
Trend reports and futures scenarios from think tanks and international organisations, alongside media reporting, academic literature and a lighter read of what people were actually saying to each other online. Breadth is cheap at this stage and worth paying for.
02
Prioritisation
We narrowed the inputs against what the education system has said it is aiming for: its stated aims for what a graduate should be able to do, its published competency frameworks and the curriculum changes already announced. This keeps a global scan answerable to local priorities.
03
Application
Four scenarios worth preparing for, and from those, two broad sets of experiments a school or system might rehearse. An output that is not something someone could do next term is not an output.
Inputs
Institutions
UNESCO, OECD, ADB, ASEAN, World Bank, WEF
Think tanks
Brookings, HolonIQ, Horizon, WISE, Educause
Media reports
EdSurge, The Economist, IFTF
Academic literature
SSRN, arXiv, white papers
Social pulse
LinkedIn, TikTok, Facebook, Instagram
Scanning tools
Feedly, X, Meltwater, Google Alerts
Method
The six source families are uneven on purpose. They range from institutional reports with a two-year lag to social platforms with a two-day one. Institutional publications tell you what has already been agreed is important, which is useful and always late. Social platforms tell you what is being felt before anyone has written it down, which is early and mostly noise.
Neither substitutes for the other. A scan drawing only on the respectable half of that list will reliably produce a picture of the recent past with the dates changed. A scan drawing only on the fast half will mistake a fortnight for a trend.
The narrowing takes four moves: collect, tag and categorise, interpret, prioritise. The first is easy and the second is tedious. The third is where the thinking happens, and the fourth is where the exercise either earns its credibility or quietly loses it.
We prioritised against the system’s stated aims. That is defensible, and it has a cost. Judging a signal by its fit with what the system has already committed to filters out the signals most worth attending to: the ones that would require a commitment to be revisited. I do not know a clean way around this. The partial remedy is to keep a visible list of what was discarded and why, so the filter can be argued with later rather than disappearing into the process.
Signals were read through seven lenses: the familiar social, technological, economic, environmental and political set, plus demographic and pedagogical. We added those two because a scan for schools that cannot see enrolment decline or teaching practice is not a scan for schools.
The lenses do a procedural job, and they partition nothing cleanly. They force a signal to be looked at from a direction the person holding it would not have chosen. Almost every useful reading in the exercise came from a signal being made to answer a lens it did not obviously belong to.
Outputs
Social
Mental health, learner expectations and family education choices
Technological
AI tutors, extended-reality classrooms and verifiable credentials
Economic
Reskilling trends, education technology investment and affordability
Environmental
Climate education, green campuses and disaster preparedness
Political
AI governance, curriculum reform and data sovereignty
Demographic
Ageing populations, migration and declining enrolments
Pedagogical
Competency-based systems and teacher–AI collaboration
What survived
Each signal the scan surfaced was read twice, once in its best light and once in its worst. A signal you can only see one way is a signal you have already made up your mind about.
Global classrooms with porous pathways, and the freedom to choose curricula that cross borders. Wide diversity in how teachers use technology for learning.
Learners who accept whatever the personalisation offers them, and lose the curiosity, compassion, imagination and empathy that were never in the recommendation.
Human purpose and agency complemented rather than replaced by machine intellect, as spatial and quantum computing, human augmentation, holographic conferencing and neural interfaces reach scale.
Overreliance to the point of being unable to function or communicate without it, an unhealthy attachment to the digital world and an equally extreme backlash against technology being in every part of life.
Cultural richness and increased empathy, producing a vibrant mix of ideas, traditions and experiences that feeds creativity and innovation.
Social diversity exploited by populist movements, producing political tension, polarisation and instability.
Rapidly evolving roles that reward innovation and creative problem-solving, with new skills and technologies raising productivity.
Education systems that cannot keep pace, a widening skills gap and worsening income inequality between those who adapt quickly and those who fall behind.
Rapid advances in clean energy, carbon capture and sustainable technology, and the green industries and jobs that follow.
More frequent and more severe weather disasters, biodiversity loss, resource scarcity, mass migration and economic instability.
Stronger international alliances and more effective global governance, with innovation pulled into cybersecurity, energy independence and space.
Conflict, economic instability and rising authoritarianism, with civil liberties curtailed and the environment degraded further by competition for resources.
Local
Read against local considerations, three socio-cultural developments held across all four scenarios. That makes them worth acting on now, whichever future arrives.
The need to reinvent Singapore amid global political, economic and technological volatility
Expanded curriculum demands, among them critical discernment and global literacy
A desire for a we-first society and a broader definition of success, amid me-first sentiment
Greater emphasis on collective contribution, and a wider set of measures for assessing and reflecting success
The challenge of increasing operational demands with shrinking resources
Preparedness for frequent disruption from climate change and from black swan events
Technology
Societal signals answer what people will want. They say nothing about what will be possible. A parallel scan of emerging technologies ran on its own criteria.
Shortlisting criteria
Five surfaced as worth watching: spatial computing, large language models, neurotechnology, edge computing and quantum computing.
Both of these are the part of the scan I would push back on hardest, and I say so on the immersive domain page as well. Whether a technology is ready to be integrated at scale and whether it should be are separate questions, and the scan that answered the first does not answer the second. Neurotechnology in particular arrives with consent and data questions that no educational use case above disposes of.
Technologies that blend the physical and digital worlds, either by overlaying digital information on the physical environment or by building a virtual environment around it.
Educational use cases
The study of the nervous system in order to monitor, analyse and modulate brain activity, using neural interfaces that read from or write to the central, peripheral or autonomic nervous system.
Educational use cases
Unsettled
Collection is the easy half. Everything below is about the half that decides what the exercise is worth.