SilverWatch by The Little Red Lens
Inspiration
Social isolation is one of the hardest problems to see.
A neighbourhood may appear well supported, yet still contain many older residents who live alone, face socioeconomic vulnerability, or have limited access to nearby community support. At the same time, an area that looks manageable today may become a much larger challenge as its population ages.
We started with one question:
What if Singapore could see these gaps before they become crises?
That question inspired The Little Red Lens.
Our concept, SilverWatch, is designed to bring together demographic data, living arrangements, socioeconomic vulnerability and social-care infrastructure into one neighbourhood-level decision-support platform.
Instead of designing another dashboard that only shows where elderly residents live, we wanted SilverWatch to answer three more useful questions:
FOCUS
Where may support be most needed today?
FORESIGHT
Which neighbourhoods could become tomorrow's ageing hotspots?
SHIFT
Where could additional community resources make the greatest difference?
The idea is to help social workers, volunteer coordinators and community planners move from seeing a hotspot to understanding why it matters and what action could potentially follow.
What it does
SilverWatch is designed as a neighbourhood intelligence platform for identifying current and emerging social isolation risk across Singapore at HDB town and planning-area level.
The platform would combine public datasets covering:
- elderly and future-elderly population
- seniors living alone
- household and socioeconomic vulnerability
- Active Ageing Centres and Senior Activity Centres
- community support coverage
Rather than displaying only a red hotspot on a map, each prioritised area would also explain the factors contributing to its risk.
For example, an area could be highlighted because it has a high concentration of seniors, a relatively high proportion of older residents living alone, socioeconomic vulnerability and weaker Active Ageing Centre coverage relative to its elderly population.
Our guiding principle is:
Every hotspot should have a reason.
The experience is structured around three lenses.
FOCUS
FOCUS is designed to show areas where several current vulnerability and support-gap signals overlap.
A user could quickly see which neighbourhoods require closer attention and understand the indicators contributing to their ranking.
FORESIGHT
FORESIGHT is designed to look beyond current risk.
An area may not rank highly today, but its elderly population could be growing quickly while support infrastructure remains relatively limited.
By incorporating future-elderly population and ageing trends, SilverWatch could highlight neighbourhoods where pressure may emerge before those areas become critical.
SHIFT
SHIFT is designed as a resource-planning layer.
A planner could test scenarios such as placing an additional Active Ageing Centre in an underserved area and compare how different locations may reduce support gaps.
This allows SilverWatch to move beyond:
Where is the problem?
towards:
What could we change, and where might that change matter most?
How we plan to build it
SilverWatch is designed around an end-to-end Databricks architecture.
We plan to use Singapore public datasets including:
- Key Indicators on the Elderly from SingStat
- HDB Elderly & Future-Elderly Population
- Resident Population Aged 65+ by Living Arrangements from SingStat
- Senior Activity Centres and Active Ageing Centres from MSF
- Households Assisted Through ComCare Schemes from MSF
These datasets each describe a different part of the problem.
Demographic data shows where older residents live and how neighbourhoods are ageing.
Living-arrangement and social-assistance data provide indicators of possible vulnerability.
Community infrastructure data shows where support currently exists.
Our proposed architecture would connect these datasets through a common HDB town and planning-area geospatial layer.
Proposed Databricks workflow
Data would be ingested through Databricks Lakeflow and stored as Delta tables.
The datasets would then be cleaned and standardised so information published at different geographic levels can be compared through a common planning-area layer.
From this unified dataset, we plan to derive indicators such as:
- elderly population density
- seniors-living-alone rate
- future-elderly growth
- ageing trajectory
- socioeconomic vulnerability
- Active Ageing Centres per elderly resident
- elderly population per centre
- distance to nearby support
- relative support coverage gap
For FOCUS, we plan to create an explainable composite vulnerability score.
Rather than using a black-box model to claim that a particular senior is socially isolated, the score would identify neighbourhoods where several risk signals overlap.
For FORESIGHT, we plan to explore machine-learning models that use demographic trends to identify areas facing increasing ageing pressure. MLflow could be used to compare and track those models.
Databricks SQL and AI/BI would support the heatmap, neighbourhood rankings and analytical views.
Genie could provide a natural-language interface so non-technical users could ask questions such as:
Which planning areas are ageing quickly but have relatively low AAC coverage?
Unity Catalog would provide governance across the datasets and analytical assets.
Challenges we considered during the design
Defining social isolation responsibly
One of the biggest design challenges is deciding what "social isolation risk" actually means.
There is no single public dataset that directly identifies whether a senior is socially isolated.
Many available indicators are only proxies.
For example, a senior living alone is not necessarily lonely, while a senior living with family is not necessarily socially connected.
Because of this, we do not want SilverWatch to make individual-level claims that the available data cannot support.
Our proposed approach is therefore neighbourhood-level risk identification rather than individual diagnosis.
SilverWatch would highlight areas where multiple vulnerability and support-gap indicators occur together.
Combining datasets at different geographic levels
Another challenge is that the available datasets may be published at different levels, including HDB blocks, towns and planning areas.
To compare demographic demand with community support meaningfully, these datasets need to be aligned to a consistent spatial unit.
We therefore see the geospatial linking layer as a core part of the proposed solution rather than simply a visualisation step.
Avoiding a generic dashboard
We also wanted to avoid designing a dashboard filled with indicators but lacking a clear decision pathway.
The proposed user experience is therefore centred around five questions:
- Where should I look?
- Why has this neighbourhood been highlighted?
- Is the pressure already present or still emerging?
- What support currently exists nearby?
- What intervention could potentially reduce the gap?
This thinking led to our FOCUS, FORESIGHT and SHIFT framework.
What we learned
Even at the design stage, one of our main lessons was that responsible AI does not mean using machine learning for every part of the problem.
For current vulnerability, interpretability may be more valuable than model complexity.
A social worker or volunteer coordinator should be able to understand why an area has been prioritised.
We also learned that individual datasets become much more useful when viewed together geographically.
An elderly population count tells one story.
An Active Ageing Centre location tells another.
But placing demographic demand, vulnerability indicators and support infrastructure on the same geographical layer creates a much more actionable view.
Most importantly, we learned that the value of SilverWatch should not be measured by the number of charts it contains.
The real question is whether the platform could help someone make a better decision.
That became the design principle behind our concept:
Do not just show the hotspot. Explain it, anticipate what comes next, and help someone respond to it.
What's next for The Little Red Lens
If selected for Round 2, our next step would be to turn the concept into a working prototype on Databricks.
We would first build the core data pipeline and harmonise the public datasets into a common planning-area layer.
We would then develop the FOCUS risk index, followed by the FORESIGHT ageing model and the SHIFT resource-allocation simulator.
The prototype would aim to demonstrate a simple end-to-end journey:
Singapore heatmap → priority area → reason for risk → future pressure → potential intervention.
Beyond the initial prototype, we would like to explore:
- travel-time accessibility to community facilities
- richer resource-allocation scenarios
- additional community and healthcare infrastructure datasets
- projections towards 2030
- local activity and programme recommendations for volunteers
- stronger model explainability and scenario comparison
Longer term, The Little Red Lens could potentially extend beyond ageing.
The same framework could be adapted to other community challenges where vulnerability is unevenly distributed across Singapore:
FOCUS on what is happening.
FORESIGHT into what may happen next.
SHIFT resources towards where they may matter most.
Sometimes the most important problems are not the ones we cannot solve.
They are the ones we have not learned how to see yet.
Log in or sign up for Devpost to join the conversation.