Generative Urbanism in India

Begin with the opportunity, because it is genuine and large. India is in the middle of one of the greatest urbanisations in human history: hundreds of millions of people moving to or being born into cities over a few decades, vast new urban extensions and whole new city projects, and an existing urban fabric under immense pressure from growth, infrastructure deficits and climate stress. Planning at this speed and scale, with limited time and resources, is exactly the kind of staggeringly complex problem where computation earns its place - not to decide, but to help planners handle complexity, test scenarios, and explore far more options than a hand could draw. Where the questions are 'how does this network perform', 'where does this growth put pressure', 'what does this density do to daylight and movement', computational analysis is genuinely useful.

India also has an unusual asset: a large, capable and globally connected IT and computational sector, and a growing community of designers, planners and researchers fluent in data and modelling. The talent that could apply generative and parametric methods to Indian cities exists at home, in strength - a real advantage over contexts where such capacity must be imported. Alongside it sit ambitious public programmes - the Smart Cities Mission and related digital-governance and urban-data efforts - that have pushed data, sensing and technology-led planning onto the agenda of Indian cities in a way that creates both infrastructure and appetite for computational methods.

So the opportunity is not hypothetical: the scale demands tools that handle complexity, the talent to build and wield them is present, and public programmes have opened the door. Used well - for analysis, for exploring options, for grounding argument in evidence, for making the trade-offs of a plan visible to the public - computational urbanism could genuinely help Indian cities plan more knowingly under enormous pressure. This is why the course refuses to treat computation as a foreign luxury: in India its potential value is arguably higher than almost anywhere, precisely because the urban challenge is so large. But 'used well' is carrying the entire weight of that sentence - and the next sections are about how easily, in the Indian context specifically, it is used badly.

Now the first and sharpest danger, because it is specific to contexts like India's: a very large share of the Indian city is informal and organic, and it rarely fits a computational model's categories. The dense old core of an Indian city, the vast informal settlements that house a huge fraction of the urban population, the street that is home and shop and workshop at once, the livelihood no zoning name captures, the lane that appears on no official map - this is not a marginal exception to the 'real' planned city; in much of urban India it *is* the city, where most people actually live and work. And it is precisely the part a model is least able to represent.

The problem is structural, not incidental. Computational methods run on data and categories - a plot has an owner, a building has a use, a street has a width, a parcel has a zoning - and the informal city fits none of them cleanly: tenure is unclear or undocumented, uses are mixed and shifting, the fabric is too fine-grained and too undocumented for the dataset. So a naive generative or optimization model does one of two things, both disastrous. It cannot see the informal city at all - it is a blank on the map, an empty parcel awaiting 'development' - or it sees it only as a problem to be cleared, a low-value inefficiency the optimization improves by removing. Either way, the model's blindness becomes a recommendation to erase the homes and livelihoods of the most vulnerable people in the city, with the false authority of an objective output.

This is the optimization trap in its most consequential Indian form. The unmeasurable and the unmapped are not abstractions here; they are hundreds of millions of people whose entire urban existence is invisible to the categories a model uses. A generated masterplan that scores beautifully may have achieved its score partly by not counting them - and 'the algorithm says redevelop this parcel' can launder a decision to displace a community into a technical result. The urbanist working in India must therefore treat the informal and organic city as the first thing to defend, not the last to consider: insisting that what the model cannot see is real and often more important than what it can, refusing to accept a blank on the map as empty ground, and keeping the people the data omits at the centre of every decision the data informs.

The second danger is historical and cultural: India has a long, cautionary history of top-down, technocratic planning, and computational urbanism is a powerful new way to repeat its mistakes with a fresh gloss of objectivity. From colonial-era planning that imposed abstract order on living cities, through the great modernist experiment of Chandigarh - a city drawn whole by Le Corbusier as a rational plan, admired and also long critiqued for how its abstract order sat against how Indians actually use urban space - to elements of contemporary smart-city and redevelopment schemes, the recurring pattern is a plan conceived by experts and imposed from above, coherent on paper and often at odds with the fine-grained, adaptive, informal life of the real city. This is exactly the drawn city's characteristic failure, and it has a specific Indian lineage that any computational urbanist here should know.

The danger is that computation automates this tendency and gilds it. A generative masterplan is top-down planning at a new scale and speed, and it arrives wearing the authority of data and algorithm - making the old technocratic move harder to argue with, because 'the model optimized it' sounds more objective than 'the planner decided it', even though the same abstract order is being imposed on the same living city. Learning India's planning history is not antiquarianism; it is the direct warning that the tool in your hands is the latest and most powerful instrument for a mistake this country has made repeatedly, and that objectivity is exactly the disguise under which it will be made again.

The third danger sharpens the first two: deep inequality, and weak protections for the poor, make the question of who is optimized for and who is displaced a matter of survival rather than convenience. In a context of stark disparities in wealth, tenure security and political voice, a computational masterplan's implicit answer to 'whose city is this?' has brutal consequences. The comfort, access and land value the model rewards will tend to be that of those who commissioned it and those the data represents - the formal, the propertied, the counted - while the cost falls on those the data omits and the process excludes. Displacement that a metric registers as an efficiency gain can be, for an informal community, the loss of home, livelihood and social fabric at once. So equity in the Indian context is not a soft consideration layered on top of computational urbanism; it is the first-order question, and the urbanist's duty to defend the equitable and keep the binding choices democratic is at its most urgent precisely here.

Holding the opportunity and the dangers together, what is the honest path for generative urbanism in India? Not techno-optimism that reaches for computation as a shortcut past the messy, political, informal reality of Indian cities - that is how the harm happens. And not a refusal that dismisses powerful tools a country urbanising this fast genuinely needs. The path is a rooted, hopeful, critically-aware practice: hopeful because computation really can help India plan more knowingly under immense pressure; rooted because it stays anchored in the actual Indian city, informal and organic and unequal as it is, rather than the tidy abstraction the model prefers; and critically aware because it treats the field's dangers as first-order, not footnotes.

In practice this means specific disciplines. Point computation at analysis and exploration - understanding Indian cities, testing scenarios, making trade-offs visible - and not at deciding them. Treat the informal and organic city as the first thing to defend: assume the model cannot see it, go and find what the data omits, and never accept a blank on the map as empty ground. Know the cautionary history well enough to recognise when a generated plan is the old top-down move in new clothing, and refuse to let 'the algorithm says' put an objective gloss on a political act of displacement. Put equity first because the inequality is deep and the protections thin, asking always who gains, who bears the cost, and who is rendered unseen. And keep the binding choices where India's own frameworks place them - the statutory master-plan and development-plan process, the applicable development-control regulations and the National Building Code of India, the planning authority, and above all the affected communities and the democratic and participatory process.

Done this way, computation serves the Indian city rather than overwriting it. The hope is real and worth holding: a generation of Indian urbanists, fluent in these methods and clear-eyed about their limits, using computation to handle genuine complexity while defending the fine-grained, informal, human city that no model can capture - and insisting, every time, that the decisions about who India's cities are for stay democratic, just, and rooted in the lives the data leaves out. That is not a compromise between opportunity and danger; it is the only way to realise the opportunity without inflicting the danger. India, of all places, needs its computational urbanists to be exactly this - powerful with the tools, and never captured by them.