The Role of the Planner

Start with the work itself, stripped of the anxiety. When computation takes over the analysis - reading data, generating options, testing scenarios, checking the DCR - it removes a large amount of technical labour from the planner's desk. The naive reading is that this hollows out the profession. The correct reading is that it *exposes the core*, because the labour it removes was never the essence of planning; it was the overhead. What remains when the overhead is gone is precisely what the planner has always uniquely owned, and it is not automatable in principle, not merely in current practice.

The first thing the planner owns is judgement - specifically, judgement across incommensurable things. A model optimizes on a common scale: it can trade daylight against density because it has reduced both to numbers. But the real choices in planning are between things that share no scale - the value of an old neighbourhood's community against the housing a redevelopment would provide, the convenience of a road against the severance it causes. No optimization can make that trade, because there is no true exchange rate between them; only human judgement can weigh them, and weighing them *is* the job. The second thing is synthesis - holding the whole of a place at once: its history, its politics, its feel, its people, its future, integrated into a coherent sense of what should be done. A model decomposes a city into measurable variables; the planner must recompose it into a living whole, and the recomposition is a human act of understanding no decomposition can perform.

The third thing is meaning and values - the questions that precede any optimization and that computation cannot even pose: what is this city *for*? Whom must it serve? What kind of life should it make possible? These are not findings; they are commitments, and they belong to people. A model can tell you how to reach a goal efficiently; it can never tell you which goal is right, because 'right' is a value, not a variable. So the planner's core turns out to be everything the model most conspicuously lacks: judgement across the incommensurable, synthesis of the whole, and the setting of values and ends. Computation doing the analysis does not threaten this core. It hands the planner back the time and the clarity to do it well - which is why the more powerful computation becomes, the more, not less, the distinctly human work of the planner matters.

The second great thing the planner owns is mediation, and to see why it is irreducibly human you have to accept an uncomfortable truth: planning is not a technical activity with political side-effects; it is a *political* activity with technical inputs. Every real plan distributes winners and losers. Densify a district and some gain homes while others lose light and quiet; route a road and some gain access while others lose their street; rezone a quarter and land values shift, enriching some and displacing others. There is no plan that is good for everyone, and pretending otherwise is the first dishonesty of technocratic planning. Because interests genuinely conflict, someone must *mediate* them - balance them, negotiate between them, decide whose claim yields and whose prevails, and hold the result together as something a community can live with.

This mediation is exactly what a model cannot do, and not for want of sophistication. A model can only mediate conflicting interests if you first tell it how much each interest weighs relative to the others - and that weighting *is* the political decision, smuggled in before the optimization starts. Whoever sets the weights has already decided who wins; the model just executes it and lends it a false air of neutrality. Genuine mediation is the opposite of a fixed weighting: it is a living process of listening, negotiating, trading, and finding a settlement that the affected parties can accept as legitimate even when it is not what any of them wanted. It requires standing, so that parties accept the mediator's authority; it requires accountability, so the settlement can be challenged and owned; and it requires the human capacity to understand interests that were never quantified and to broker between people, not variables. None of these is a computational property.

So the planner is, unavoidably, a political actor - not in the partisan sense but in the deep sense of one who governs the distribution of goods and harms among a public. The danger computation introduces is the temptation to *hide* this political role behind the model: to let 'the optimization' appear to make the distributive choices so that no one has to own them. That is an abdication, and it is corrosive, because it removes the choices from contest without removing their consequences. The honest planner does the opposite - uses the model's evidence to mediate *better*, more transparently, with the trade-offs laid bare, while keeping the mediation itself visibly, accountably human. The city is a settlement between conflicting human claims. Only a human with standing and accountability can broker it.

The third thing the planner owns is a duty, and it is the one that most sharply distinguishes a planner from a mere operator of tools: the defence of the public good and the unmeasurable. A model, by its nature, sees only what has been measured and serves only what it was told to optimize. That means two vast territories fall permanently outside its vision - the *public good* that no private interest will speak for, and the *unmeasurable* qualities that no metric can hold - and someone must actively defend both precisely because the model structurally cannot.

The public good is everything that belongs to everyone and to no one in particular: the coherence of the city as a whole, the interests of the poor and the unrepresented, the claims of future generations, the shared commons that a sum of private optimizations will always erode. A market or a model driven by whoever commissioned it will optimize for the powerful and the present; the planner exists, in large part, to hold the line for the powerless and the future, to be the voice in the room for the interest that has no other advocate. This is not a technical function; it is a *fiduciary* one - a duty of care to a public - and it is the moral heart of the profession.

The unmeasurable is the other territory: community, belonging, memory, meaning, the fine-grained mixture of life, the dignity of a place and its people - all the things the optimization trap sacrifices because they never entered the objective function. The planner's defence of the unmeasurable is the discipline this entire course has built toward: to insist, against the seductive precision of the metrics, that the parts of a city that make it worth living in are largely the parts the model cannot see, and to protect them *from* the optimization, not with it. In India this duty is at its most acute around the informal and organic city - the settlement the model does not represent, the livelihood it cannot quantify, the community it would 'clear' as empty land. The planner is the one who must put that invisible city on the table, speak for it against the clean logic of the plan, and refuse to let computational blindness become the plan's cruelty. Owning judgement and mediation makes the planner competent. Owning the defence of the public good and the unmeasurable makes the planner necessary - the conscience the model can never contain.

Everything so far converges on the deepest thing the planner owns, which is not a skill at all but a *position*: the planner is the accountable human who answers for the city. When a plan harms someone, when a decision must be justified to a citizen, a court, an elected body or the public, there must be a person or an institution that stands behind it and can be held to account - that can be questioned, blamed, made to explain, and ultimately replaced. This accountability is not a bureaucratic detail; it is the entire basis on which anyone is entitled to exercise power over others in a democracy. And it is the one thing a model can never provide, because a model cannot answer. It cannot be summoned to explain itself to a displaced family, cannot be voted out, cannot bear responsibility, cannot care about the consequences it produces. Accountability requires a subject who can be held answerable, and an algorithm is not such a subject.

This is why 'the algorithm says' is the most dangerous phrase in the field, and why this lesson ends on it. It is an attempt to exercise accountable power without an accountable person - to make a binding choice about people's lives while ensuring no one can be held responsible for it. It launders a human decision (someone chose the model, the data, the objective, the deployment) into an ownerless technical output, and an ownerless decision is precisely a decision no democracy should accept. The planner's ultimate role, then, is to *refuse the chair to the algorithm* - to insist that behind every use of computation stands a human who owns the choice, carries the responsibility, and can be called to account. The planner does not compete with the model on analysis; the planner sits in the chair the model must never occupy.

So the whole anxiety inverts. Computation does not replace the planner; it makes the planner's irreducibly human core - judgement, values, mediation, the defence of the public good and the unmeasurable, and above all accountability - more visible and more necessary than ever. The competent computational urbanist of the future is not a technician who defers to the model, but an accountable human who wields the model powerfully for what it is good at while owning, personally and professionally, everything it decides. In the Indian frame and everywhere, the binding choices about a city's future stay with the planner as accountable human, the planning authority, the democratic and participatory process, the affected communities, and the law - and the algorithm, however brilliant, never takes the chair.