The Staircase and the Platform

Yesterday, The Guardian returned to Finland’s remarkable Housing First experiment. Since 2008, the country has treated permanent housing not as the final reward for successfully completing treatment, rehabilitation, and temporary accommodation, but as the starting point: a model that »focuses on housing as a fundamental human right rather than a reward«. Finland has seen striking results: between 2012 and 2023, homelessness fell by more than half. Now, after cuts to benefits and social services, the trend has begun to reverse. The experiment was never a magic cure. It was a different way of arranging the problem.¹

»The home isn’t a staircase: it’s a platform«

The interesting thing about Housing First isn’t just that it works. It changes the order in which the problem is attacked. We tend to imagine difficult problems as staircases – first fix A, then B, then C, and eventually you arrive at the desired outcome. Often this is exactly right: when building a house, you need foundations before you can put up the walls. But sometimes the staircase contains a trap: the thing you’re supposed to achieve at the end is also the thing that would make the other steps possible. A person without a home may need treatment, employment, financial stability, and better health. But trying to achieve all of this while sleeping on the street is rather like asking someone to repair a car while driving it. Giving them a home won’t make the other problems disappear, but it will dramatically expand what they can do about their situation. The home isn’t a staircase: it’s a platform.

There is a useful analogy in artificial intelligence. Hierarchical Task Network planning breaks complex goals into smaller tasks, with conditions determining which actions can happen and in what order. The interesting question perhaps isn’t what task comes next, but what becomes possible if we change the state of the system.²

A staircase is infrastructure designed around a sequence. A platform is infrastructure designed around possibility.

We build roads because it would be absurd to ask each person to construct a little piece of road before making a journey. The road is a shared platform: once it exists, thousands of people can use it in ways its builders never had to predict. Scientific infrastructure works in much the same way. A telescope doesn’t solve an astronomical problem; it makes thousands of possible questions easier to ask. A particle accelerator doesn’t tell researchers what to discover; it expands the territory in which discovery can happen. Platforms are particularly valuable when we don’t know exactly what will be needed.

The distinction also suggests a different way of thinking about public spending. We normally ask what a new intervention will cost. We rarely ask what the existing system costs because that intervention doesn’t exist. How much does homelessness cost through emergency healthcare, shelters, and crisis services? How much does preventable illness cost through hospitals and lost productivity? How much does congestion cost through roads, pollution, and time? This is what we might call counterfactual budgeting: costing not only the proposed solution, but the world in which we don’t build it.

The point is not to manufacture a precise number where there isn’t one, but to make the assumptions visible: what the intervention costs, what it might save, and when the two curves might cross. Sometimes the platform will cost more than the problem it solves. Sometimes it will cost less. The difficulty is that governments are organised into departments, and problems are not. The department that pays for housing may not be the one that saves money on healthcare, policing, or emergency accommodation. One budget gets worse. The system gets better.

This is why platforms are politically awkward. Their costs are immediate. Their benefits are distributed, delayed, and uncertain. Finland’s own platform ran into exactly this asymmetry: the savings it generated were scattered across hospitals, police, and crisis services, while the cuts that undid it landed in a single, visible budget line. Yet we already know how to build these things: roads, schools, libraries, laboratories, electricity grids.

Perhaps the next question is what else deserves to be treated as infrastructure. Should everyone over 40 get a personal trainer, the way we already treat childhood vaccination as public infrastructure rather than a private choice? Should a phone, broadband connection and bank account be considered basic infrastructure rather than consumer goods? Should we build scientific platforms before we know what discoveries they will enable? Wherever millions of people repeatedly encounter the same constraint, it may be cheaper to build a platform than to ask them to overcome it individually, one staircase at a time.

The Finnish lesson isn’t simply that we should house homeless people. It’s that sometimes the most intelligent way to solve a problem is to stop attacking the problem.

Build the platform first.

References