The Parametric Model

Every parametric model, whatever the software, has the same three-part anatomy, and seeing it plainly demystifies the whole enterprise. Inputs are the parameters you chose to expose - block size, street width, setback, FSI, height, orientation and the rest - each a knob with a range. Logic is the set of relationships that convert those inputs into a design: rules like "lay a street grid at this spacing, bound blocks with it, subdivide each block into plots of this depth, place a building set back this far and massed to this coverage and height". Outputs are what the logic produces: the plan itself as geometry, plus derived metrics you compute from it - the resulting density, the floor area, daylight hours, the count of dwellings, a walkability estimate. The flow is one-directional and mechanical: inputs feed logic, logic produces outputs, and nothing in the output was decided anywhere but in the inputs and the logic.

That last sentence is the whole discipline in miniature, and it cuts two ways. On the good side, it means the model is *transparent in principle*: every feature of the output traces to a parameter you set or a relationship you wrote, so there is no magic, nothing the model "knows" that you did not put in. On the sobering side, it means the model can only ever be as good as its inputs and logic - the old computing truth, garbage in, garbage out, applies with full force. A beautiful, precise, instantly re-forming plan built on a wrong relationship or an inappropriate parameter range is a beautiful, precise, instantly re-forming piece of nonsense, and its very polish makes the nonsense harder to spot.

The metrics deserve special wariness, because they are where the model feels most authoritative. When the output panel reports "density 240 dwellings per hectare, average daylight 4.2 hours, walkability 78", those numbers look like measurements of a real place. They are not. They are *consequences of your assumptions*, computed to spurious precision - the daylight figure assumes a sky model and a massing rule you chose, the walkability score assumes a definition of walkability someone encoded. Treat outputs, and especially metric outputs, as "what this model says follows from these assumptions", never as "what is true of the city". The anatomy is simple; the humility it demands is the hard part.

What makes a parametric model re-form rather than break is its associative structure: elements are defined in terms of one another instead of pinned to absolute coordinates. Think of it as a dependency graph. The street grid depends on the street-spacing parameter; blocks depend on the grid; plots depend on the blocks and the plot-depth rule; buildings depend on the plots, the setback, the coverage and the height; the density metric depends on the buildings and their unit mix. Each element points to the things it is built from. When you change one input - say block size - the change flows downstream along these dependencies: street spacing updates, blocks resize, plots re-subdivide, buildings replace, density recomputes, all automatically and all consistent, because each element simply rebuilds itself from its (now-changed) parents.

This is a genuinely powerful way to hold a design, and it is why parametric models feel almost alive. You are not maintaining a plan; you are maintaining the *reasons* for a plan, and letting the current plan fall out of them. It also localises your thinking: to change how density responds to the transit stop, you edit one relationship, and every plan the model can produce inherits the change. Compared with hand-drafting, where consistency is a constant manual chore and a single forgotten edit leaves a self-contradicting plan, associativity gives you consistency for free, guaranteed by construction.

But the dependency graph is also where assumptions hide most effectively, and this is the subtle danger to internalise. Every edge in the graph is a claim: "plots depend on blocks" assumes plots are carved from blocks bounded by streets - which quietly forbids the plot that fronts a courtyard, the informal cluster that grew without a block, the fabric that is not a grid at all. The relationships that give you consistency are the same relationships that fix your assumptions into the bones of every variant. You cannot turn a knob to escape them, because they are not knobs; they are the wiring. So a parametric model's real commitments live not in its adjustable parameters but in its *fixed relationships* - the graph you cannot see in any single plan. Reading a model critically means reading that graph: not "what can I tune?" but "what did the wiring already decide, for every plan this model will ever make?".

Because a parametric model is an argument about how a city works, dressed as geometry, the difference between a trustworthy model and a dangerous one is not sophistication but *discipline*. A good model is, first, legible: someone other than its author can open it and understand what the parameters are, what the relationships assume, and where the numbers came from. Cryptic models that only their maker can read are not clever; they are unaccountable, and unaccountable is exactly what an instrument of public consequence must never be. Document the parameters, their ranges and their sources - ideally the specific regulation each came from - and state the key relationships in plain language beside the geometry.

Second, a good model is honest about its assumptions. Every model rests on simplifications; the sin is not making them but hiding them. A disciplined modeller keeps an explicit list of what the model assumes and therefore cannot represent - "assumes rectangular blocks; assumes plots face streets; assumes uniform density within a zone; does not represent informal settlement; daylight uses this sky model" - and hands that list over with every result. This single habit does more to prevent harm than any amount of computational power, because it lets others challenge the assumptions rather than swallow the output. Third, a good model is bounded and validated: parameter ranges reflect real regulatory and physical limits rather than arbitrary sliders, and outputs are sanity-checked against reality - does this density match comparable real neighbourhoods, does this daylight figure survive a hand check on a known case? A model that has never been reconciled with a real place is a hypothesis, not evidence.

Finally, a good model is scoped to a question. The temptation is always to model more - more parameters, more coupled systems, a digital everything. But a sprawling model is less legible, less validated and more likely to hide a fatal assumption, while adding a false impression of completeness. The disciplined move is to build the smallest model that genuinely answers the design question at hand and to state clearly what it deliberately leaves out. Across all four disciplines runs one principle: the model exists to *inform and to open questions for human judgement*, not to settle them. Build it so that its assumptions can be argued with, its numbers checked, and its silences seen - because a model whose reasoning cannot be interrogated has no business influencing a public decision about a city.

Now the honest heart of the lesson. The gravest risk of a parametric model is not that it computes something wrong - a wrong number can be caught - but that its *assumptions silently become the plan*, and then the plan silently becomes the city, with no one ever having decided the assumptions on purpose. Recall that the model's real commitments live in its fixed relationships, invisible in any single output. When such a model drives a real masterplan, everything those relationships assumed - that the city is a grid of street-facing plots, that density is uniform, that a street is for movement, that what is not in the model is not there - passes into the built world without ever being stated, debated or approved. The most important decisions about the plan were made by whoever wired the graph, often a technician optimising for legibility or performance, and they were made once, invisibly, for every plan the model would ever produce.

This is how a parametric model launders assumption into fact. The output is quantified, consistent and re-forms at a touch, so it carries the authority of computation - it looks discovered rather than assumed. Officials and the public see a clean plan and precise metrics, not the buried claim that the informal settlement occupying part of the site simply does not exist because the model has no category for it. The false objectivity is not a bug the modeller intended; it is the natural effect of a machine that hides its premises inside relationships and shows only polished results. In the Indian context the stakes are stark: a model built on formal, rectangular, street-facing assumptions can render the living informal city invisible, and a plan grown from it can then erase real homes and livelihoods while appearing merely to "follow the analysis".

The discipline that answers this danger is the one this whole module has been building toward. Treat every parametric model as an argument to be interrogated, not an oracle to be consulted. Insist that its assumptions and silences be listed and tabled wherever it informs a decision. Keep its role to exploring and testing - showing what follows from stated premises - and never let the clean output stand in for the messy, contested, human reality it simplified. And hold the binding line without exception: the choices a parametric model can influence but must never make - what a city builds, at what density, for whom, and whose existing fabric is protected or displaced - belong to the planning authority, the democratic and participatory process, the affected communities and the governing planning law and development-control regulations. The model informs the argument; people, accountably, decide the city.