Why We Still Need the Wrong Map

One of the most useful maps ever produced takes considerable liberties with the place it describes. Distances stretch and shrink. Railway lines straighten themselves into arrangements their engineers would not recognise. Central London expands, the suburbs retreat, and the Thames receives instructions to behave more tidily.

The London Underground map is an extraordinary achievement in being wrong about the right things. Give a visitor a sufficiently accurate drawing of the tunnels and you may improve their geographical knowledge while substantially reducing their chances of reaching dinner.

This is an uncomfortable proposition in a culture that treats more information as an automatic improvement. Sometimes understanding begins when somebody has the nerve to leave most of it out.

London, with the inconvenient parts removed

Harry Beck’s diagram first appeared in 1933. As London Museum explains, his design departed from geographical representation, arranging the network through simplified lines and clearer relationships between stations.

The passenger’s problem was practical. Which line should I take? Where should I change? How many stops remain before I must interrupt my increasingly confident impression that I know what I am doing?

The precise curve of the railway beneath somebody’s kitchen contributed little to those decisions. Beck gave space to the crowded centre and compressed the outlying stretches. Connections became easier to follow because physical distance stopped receiving equal consideration.

His achievement was editorial as much as cartographic. He distinguished between information that described the system and information that helped someone use it. These are related categories, but anyone who has opened a printer manual will understand that they are not identical.

There is no innocent flat map

Even a conventional geographical map cannot simply present the world without making choices. The Earth is curved. Paper is flat. Something has to give, and the paper has been remarkably firm about its position.

The US Geological Survey’s account of map projections explains the resulting compromises. Different projections preserve different properties, sacrificing some combination of shape, area, distance or direction. There is no projection that keeps everything correct everywhere.

Mercator’s projection is useful for navigation because constant-bearing routes appear as straight lines. It also greatly enlarges areas towards the poles. The familiar result is a world in which Greenland has acquired a rather inflated sense of its own importance.

That does not make Mercator a fraud. It makes it a tool whose virtues become faults when used for a different purpose. A map suitable for plotting a course is not necessarily suitable for comparing the sizes of countries.

The question is therefore more demanding than “Is this accurate?” Accurate in which respect, across what area, and for somebody trying to do what?

The map can persuade you

Useful distortion still carries a cost. Once a diagram becomes familiar, its compromises can slip out of sight. We begin to treat the arrangement on the page as a property of the place.

Research supports taking that possibility seriously. Zhan Guo’s study, Mind the Map!, published in 2011, examined London Underground route choices and found that the schematic map affected which paths passengers selected. The representation was participating in the decision.

The everyday illustration is the short journey that looks more substantial underground than it is on foot. TfL’s West End walking map puts Leicester Square and Covent Garden about four minutes apart. Depending on the traveller and circumstances, the most useful transport information may be permission not to use the transport.

A person can descend into the network, wait for a train and emerge near where they started, having followed the map quite competently. The failure was asking a diagram of railway connections to make a decision about walking.

The answer is not to abolish schematic maps. It is to know when to put one away.

A small world can reveal a large problem

Scientific models make a similar bargain. They leave things out to expose relationships that might otherwise remain buried in detail. Their simplicity can be a source of insight, provided we remember which world they describe.

Thomas Schelling’s segregation model is a striking example. As the Royal Swedish Academy of Sciences explained, relatively modest preferences for having some similar neighbours could, through successive moves, produce strongly segregated patterns.

The model helps reveal how individual choices can generate an outcome nobody explicitly selected for the whole system. It does not establish that actual segregation is adequately explained by mild preferences. Discrimination, housing policy, unequal wealth and coercion do not disappear because they were absent from the simplified world.

That is where a useful model can become an alibi. Demonstrating that a mechanism can produce an outcome is different from establishing how much it explains the outcome in front of you.

A model earns its place by making something intelligible. It does not thereby acquire jurisdiction over everything it omitted.

The dashboard is another map

The same issue appears in less distinguished settings. A project dashboard turns months of work into dates, percentages and coloured circles. This can be helpful. Nobody attending a progress meeting wants to experience the entire project in real time, although some meetings make a determined attempt.

Trouble begins when the summary’s categories become more authoritative than the work. A task is marked complete because the defined activity occurred, although the intended result remains elusive. The map shows a connection. The passenger discovers a locked gate.

Adding more detail does not automatically solve this. A dashboard can become so comprehensive that finding the important problem requires somebody to construct a smaller dashboard beside it. We have rebuilt the original difficulty at considerable licensing expense.

The better question is what decision the summary exists to support. If it helps someone spot a delayed dependency, it should make that visible. If it is supposed to establish readiness, it must distinguish activity from a result that can actually be used.

These are design judgements, not discoveries that can be delegated entirely to the data. Somebody has to decide what matters, then accept responsibility for what the representation conceals.

Keep more than one map

The temptation is to search for a final representation containing enough detail to settle every question. But different purposes call for different omissions. A walking map, a geological map and an Underground diagram can all describe London honestly while showing remarkably little of one another’s information.

Using several representations is not an admission that knowledge has failed. It is a way to avoid asking one successful simplification to perform a job it was never designed to do.

We should be particularly suspicious when a model’s limitations are acknowledged ceremonially and ignored operationally. “Of course, this is only a model” achieves little if every subsequent decision treats it as the terrain.

The wrong map remains necessary because the alternative is not perfect understanding. It is often an accumulation of detail through which we cannot find our way. What matters is choosing the distortion deliberately, checking its consequences and recognising when the question has changed.

Beck’s diagram can get you across London. Outside the station, you need to look up. The city has not agreed to arrange its streets at forty-five degrees, and the restaurant is under no obligation to be where it would make the diagram tidier.