Coding and AI classes in Hampstead, London

The learner downloads two Census 2021 tables from the Nomis API for all 751 output areas in Camden. Across 92,763 households, 59,026 have no car or van. For each small area the program works out two percentages: households with no car, and households where one person lives alone. Across Camden the averages are 63.01% and 38.61%, and the two move together, with a correlation of 0.552: areas with more people living alone tend to have fewer cars.

The ordinary method standardises each percentage and measures straight-line (Euclidean) distance from the average. The Mahalanobis distance does the same job but first accounts for the link between the two measures, so that being high on both, which is common, counts as less surprising than being high on one and low on the other.

The two rulers agree on 14 of the top 20 and disagree on the rest. One area has 65.0% of households without a car, almost exactly the Camden average, but only 10.7% living alone. Neither figure is dramatic, and the ordinary ruler ranks it 46th. The Mahalanobis distance ranks it 16th, because so few cars with so few single-person homes is a rare pairing here. The reverse also happens: an area with about 20% on both measures looks extreme to the ordinary ruler, yet low on both is exactly what the correlation predicts, so it drops out of the Mahalanobis top 20. Among the 60 areas in Hampstead Town and Frognal wards, where on average 44.38% of households have no car, 8 pass the cut-off on the ordinary ruler and 5 on the Mahalanobis one.

Knowing which ruler was used keeps Hampstead teenagers from taking an AI's "anomaly" at face value, and that habit comes from writing the code themselves. The longer argument is in why teenagers should still learn to code in 2026.