What the Model Can’t Show You#
SLOP2418 — Urban Design Studio
Week 11 · Marisol Quaye · 17 May 2027
direct readout · derived measure · worked hypothetical
Why this lecture sits here#
- Two assessments are marked. The capstone is due next Monday. This is the last week for changing the substantive position before Friday’s memo; Monday still allows final board and screenshot preparation.
- Every week since Week 1 has asked you to make a claim and support it with an output of the model — a ridership figure, a coverage radius, a land value overlay, a budget line.
- Today asks the other question, which no jury in professional practice will let you skip: what does this evidence not cover, and what did you conclude anyway.
- This is standing practice in every design discipline that models. A structural engineer knows precisely what their analysis omits and carries the list to the meeting. That list is the subject of slide 3.
- By the end of the hour you will have written the first draft of your own list. It is a required section of the capstone memo.
The engineer’s omissions list#
| A linear-elastic frame analysis does not represent | Why not | What the engineer does instead |
|---|---|---|
| Material behaviour past the elastic range | The solution method requires linearity | Capacity checked against code limits outside the model |
| Creep and long-term deflection | Not a product of the static solve | A separate serviceability calculation |
| Connection slip and fabrication tolerance | Joints are idealised as rigid or pinned | Detailing rules and factored allowances |
| Construction sequence and temporary conditions | The model solves the completed structure | A staged analysis, or a method statement |
| Residual stress from welding | Outside the constitutive assumptions | Fabrication specification and inspection |
Below the table, two lines:
- The engineer does not stop using the analysis. They use it for what it covers, and they hold the list for everything else.
- Nobody calls that scepticism. They call it competence.
You cannot verify a model of an open system#
- Oreskes, Shrader-Frechette and Belitz, in Science (1994), separate three words that get used interchangeably. Verification: demonstrating truth. Validation: demonstrating internal consistency and absence of known error. Confirmation: observing agreement between the model and the world.
- Their conclusion: numerical models of open natural systems can be confirmed, never verified. Agreement raises confidence. It does not establish that the model is right, and it never establishes that it is right for the reason you think.
- A city is an open system. So is a district.
- Practical standard for the capstone: an output of the model is evidence for a claim about the model. It is evidence for a claim about a district only to the extent that the mechanism producing it is one you can name.
- Which means every figure in your memo needs a mechanism behind it, not just a number.
Eleven weeks of the Kerrow corridor#
Derived measure: the index is reconstructed from named weekly saves, not emitted by the game.
What the curve is made of#
- In this simulation land value is a computed field over the map. It responds to service coverage, transit access, park and plaza proximity, density, and — negatively — to noise and pollution.
- Every one of those inputs is something the studio installed with public money: the spine, the trunk upgrades in Week 7, the plaza chain in Week 9, the floodway park in Week 10.
- So the curve is not a market discovering something. It is a ledger of the capital programme, expressed as a colour.
- This is a genuinely useful instrument. It is how you tell whether an investment landed, and it is why the fiscal performance argument in your Week 8 memo worked.
- Note carefully what it is not: it is a value, not a price. Nobody in the model offers it, nobody accepts it, and nobody holds the asset it is a value of.
114 points of uplift, accruing to nobody#
- The corridor’s land value rose 114 index points in eleven weeks, on the back of investment the district’s rates paid for.
- Henry George, Progress and Poverty (1879): the increase in land value produced by the growth and investment of the community is not produced by the landholder, and George’s argument is that the community should recover it. The whole field of land-value capture — betterment levies, developer contributions, land readjustment — descends from that observation.
- Outside the model there is always an answer to “who got the 114 points”. Someone held the land on the day the spine was announced and someone held it after. That transfer is the single most consequential distributive fact about a corridor upzoning.
- Inside the model there is no answer, because there is no landholder entity. Land value is an attribute of a tile. It cannot be owned, borrowed against, assembled, optioned, sat on, or sold.
- Consequence for your plan: the model cannot tell you whether your corridor strategy transferred public investment into private hands, because it has no hands. If your memo claims the corridor was good value for the public, that claim is outside your evidence and has to be argued another way.
What the model does show about who left#
the model is not silent about households. It is thin about them. Here is exactly how thin.
the Kerrow corridor, a populated Week 4 hypothetical against Week 11:
| Measure | Week 4 hypothetical | Week 11 hypothetical |
|---|---|---|
| Land value index | 100 | 214 |
| Rent index | 100 | 231 |
| Low-wealth households | 340 | 96 |
| Medium-wealth households | 210 | 288 |
| High-wealth households | 40 | 402 |
| Total households | 590 | 786 |
| Corridor livability reading | 61 | 78 |
| District homelessness figure | 0 | 34 |
These are worked hypothetical aggregates, not captured game readouts. Land value and rent are indices (Week 4 = 100); livability is a derived worksheet score. The low-wealth count falls by 244, but counts alone cannot distinguish departures from wealth-category changes. The 34 currently homeless households cannot be matched to those 244 without longitudinal records.
Livability is measured on whoever remains#
- The corridor’s livability reading went from 61 to 78 across the same Week 4–11 interval in this hypothetical. That aggregate gain requires an account of who benefits.
- It is computed over the households present when it is read. The population it describes in Week 11 is substantially not the population it described in Week 4.
- So a district can raise its livability score by improving conditions, by replacing its residents, or by any mixture of the two, and the metric returns the same number.
- Susan Fainstein, The Just City (2010), proposes three criteria for evaluating an urban plan: equity, diversity and democracy. Set them beside the studio’s three — fiscal performance, mobility performance, livability. Two of hers have no counterpart in ours, and the missing one that matters here is equity, because equity is the only one of the six that asks for whom a number improved.
- This studio is not replacing its three metrics in Week 11. It is telling you that they are three metrics, chosen, with known silences, and that naming the silence is part of using them.
Displacement has four shapes#
Marcuse’s four forms, :
| Form | What it is | Can this model count it? |
|---|---|---|
| Direct last-resident | The household in the unit when rent rises or the building goes | Partly — it can show the household leaving, and nothing about the leaving |
| Direct chain | The households pushed out earlier in the same building’s trajectory, before the visible change | No — the model retains no history of who occupied a unit |
| Exclusionary | The households who would have moved in and now cannot afford to | No — unmet demand is aggregate, and has no household attached |
| Displacement pressure | The dispossession felt by those who stay as the shops, services and networks around them change | No — there is no concept of a support that can be lost while you remain |
three of the four are invisible in any count of current residents, which is the only kind of count this model keeps.
The model has one tenure, and it has no term#
- In this simulation, a household occupies a unit and pays rent. That is the entire tenure system.
- Tenure relations it does not distinguish: ownership (freehold, mortgaged, leasehold, co-operative or community land trust), tenancy (fixed-term, periodic, public or social with allocation rules), and everything with no documented title at all (customary or communal tenure, sub-letting, room and boarding, undocumented occupation).
- Nor the instruments attached to them: lease, notice, eviction, mortgage, covenant, easement, resumption, compensation, objection, appeal.
- Why this matters for reading slide 5: tenure is the switch that decides whether that curve is a household’s windfall or a household’s notice to leave. The curve is identical in both cases. It contains no information about which one happened.
- Elinor Ostrom, Governing the Commons (1990), is the standard warning against assuming ownership comes in only two kinds. The model assumes it comes in none.
The model’s hard zero#
- Leave land unzoned and unserviced in this simulation and nothing is ever built on it, under any level of demand, for any length of time.
- That is the omission with the largest number attached to it. A very large share of the world’s urban population lives in housing built without the zoning, title or servicing that this model treats as the precondition of a building existing.
- John F. C. Turner, “Housing as a Verb”, in Turner and Fichter, Freedom to Build (1972): housing is better understood as a process residents conduct than as a product an authority delivers. The model implements only the product.
- Ananya Roy (2005): informality is not a zone outside planning but a mode of urbanisation the state produces — through what it zones, what it services, what it tolerates and what it clears. The model has the state’s instruments and none of the consequences.
- Vanessa Watson, “Seeing from the South” (2009), and her earlier account of conflicting rationalities (2003): planning practice built on the northern assumption of a serviced, titled, regulated city misreads most of the urban world, and misreads it in the same direction every time.
- So: this model cannot be used to reason about housing that arrives without permission. If your capstone’s housing argument depends on formal supply meeting all demand, that dependency is an assumption, and it is yours rather than the model’s.
One decision-maker, no room#
- In this simulation, a zoning change takes effect the moment it is drawn. There is no notification period, no submission, no hearing, no vote, no appeal, and no developer with interests of their own.
- Sherry Arnstein, “A Ladder of Citizen Participation” (1969), arranged participation on eight rungs from manipulation to citizen control. This model is not on the ladder. There is no rung for a process with one participant.
- Paul Davidoff, “Advocacy and Pluralism in Planning” (1965), argued that the planner is not a neutral technician and that a plan is one interested party’s argument among several. The model gives you exactly one plan and no other party.
- Bent Flyvbjerg, Rationality and Power (1998), on how rationality is produced by power rather than the reverse: the plan that gets built is not the one the analysis favoured, it is the one the process could carry.
- What this costs you in practice: every timeline in your capstone is a timeline with the politics removed. A corridor upzoning that takes four simulated years here would take considerably longer against an objection process, and the sequencing of your plan is the first thing that breaks when it does.
- So state your sequencing as conditional. “Stage two follows stage one” is a claim about the model. “Stage two requires stage one to have been approved” is a claim about a plan.
Comprehensive models fail by being comprehensive#
- Douglass Lee’s “Requiem for Large-Scale Models” (1973) reviewed a decade of integrated urban models and listed seven characteristic faults: hypercomprehensiveness, grossness, hungriness, wrongheadedness, complicatedness, mechanicalness, expensiveness.
- His argument was not that modelling is worthless. It was that a model built to represent everything is too coarse at every point to be right anywhere, needs data nobody can supply, and becomes complicated enough to conceal its own assumptions from the people relying on it.
- Read that list against the model we use. It is a single integrated solve over economy, traffic, services and land value. It is comprehensive by design. Lee’s list is not a description of a bad model — it is a description of this kind of model working normally.
- Theodore Porter, Trust in Numbers (1995), supplies the other half: quantification is adopted where trust is thin, because a number travels further than a judgement. That is why your jury asks for figures, and why figures are persuasive out of proportion to what they establish.
- Which is the reason the omissions list exists. The number will be believed. You are the only person in the room in a position to say what it covers.
From eight years of rate-base modelling#
“The rate-base model was never wrong about the revenue. It had no field for the tenants, and I signed the report anyway.
Knowing which field is missing is not a criticism of the model. It is the job.”
— Marisol Quaye
The tradeoff, named#
Mobility performance against livability#
The last four blocks of the Kerrow corridor. Upzone them and the spine hits its ridership target.
Livability reads as rising too, because the derived measure is computed over the households that are there afterwards.
Both numbers go up. The trade still happened. Write it down.
The omissions appendix, and where to read further#
- Required in the capstone defence memo, half a page, four headings. This is not a reflective essay. Each heading takes two or three sentences and a figure where you have one.
- Who holds the land whose value this plan raised, and what the plan does about it.
- What can the Week 4–11 aggregate counts establish about departures, and where the plan assumes they went.
- Which of the plan’s claims rest on a figure the model computes rather than a behaviour it observes.
- What a person who objected to this plan would object to, and what the plan’s answer is.
- A memo with an honest appendix is marked better than a memo with a flattering one. The jury has read a great many flattering ones.
Where to read further#
Further reading, as a plain two-column list:
- Scott, Seeing Like a State (1998)
- Marcuse, “Gentrification, Abandonment and Displacement” (1985)
- Glass, London: Aspects of Change (1964) — where the word came from
- Neil Smith, “Toward a Theory of Gentrification” (1979) — the rent gap
- Hartman, “The Right to Stay Put” (1984)
- Lees, Slater and Wyly, Gentrification (2008)
- Turner and Fichter, Freedom to Build (1972)
- Roy, “Urban Informality” (2005)
- Watson, “Seeing from the South” (2009)
- de Soto, The Mystery of Capital (2000), read against Gilbert, “The Return of the Slum: Does Language Matter?” (2007)
- Ostrom, Governing the Commons (1990)
- Lee, “Requiem for Large-Scale Models” (1973)
- Oreskes, Shrader-Frechette and Belitz (1994)
- Porter, Trust in Numbers (1995)
- Merry, The Seductions of Quantification (2016)
- Arnstein, “A Ladder of Citizen Participation” (1969)
- Davidoff, “Advocacy and Pluralism in Planning” (1965)
- Flyvbjerg, Rationality and Power (1998)
- George, Progress and Poverty (1879)
- Fainstein, The Just City (2010)
- Lefebvre, Le Droit à la ville (1968), and Harvey, “The Right to the City” (2008)