Monday, 11 November 2013

Oh crikey!...it's the rozzers!

Police.UK just published an updated crime map.


They've been hard at work over the last couple of years since the original version which was released in February 2011.

OSM. Check
MapBox.js. Check.
Leaflet. Check.
Dynamic cluster markers. Check.
Neat customisable search area. Check
Clean UI. Check

Yep, they've got all the latest tech driving the map and in respect of the UI/UX it's a clear improvement but has it improved in terms of what it offers the casual user?  I wrote a blog on the first map and pointed out a few cartographic issues with how the data was misrepresenting geography and crime...so what's changed?

Nothing. Check.

It's a new face on an old map. The data model is the same as before...crimes are aggregated to a generic locator placed on a road segment. Only when you click on the marker symbol do you get a breakdown of crime type (and only when you're zoomed well into the map). I can filter by type but that just updates the number.  Why hasn't the map been designed so I can easily 'see' the patterns? Why make me have to interpret numbers?

Why aren't the symbols scaled properly...look in the image above. 1 crime is the same sized symbol as 8. 286 only marginally larger. This is a false picture visually and perceptually. Did 286 crimes really get committed 'here' as it says when you open the popup? Of course they didn't. They occurred across an area. So what area is that and how can I see it? This is a perfect example of the Modifiable Areal Unit Problem at work. Every number in the map is a function of the size and shape of the reporting area. What are these? It's important to understand them to gauge how those arbitrary areas affect the pattern you're seeing. Small numbers appear on smaller roads and larger numbers on larger ones...so the reporting seems to have a relationship with the nearest road and of course you'd expect more crimes over longer roads covering larger areas.

And here's the sting...unless those numbers all relate to equally sized areas, or they are on an equally spaced grid the map is clearly falling foul of that age old problem of not normalizing your data to account for differently sized enumeration areas.  I have no problem with proportional symbols of totals being used as long as I can see what the pattern of the enumeration area is...but without it I have no idea how the map is being shaped whatever slick interface I have to navigate.

My previous comments about version 1 stand...the proportional symbols are weak and suggest absolute location as well as the uncertainty of the underlying enumeration geography not being visible. Why not use a proportional linear symbol for the entire road rather than suggest all crimes occur at that specific point? Better still...why not symbolize the reporting areas to give us a truer picture of crime across space? Are area maps just not sexy enough any more?

Is this a question of the data being generalized to a nearest point or is it the mapping trying to introduce some fuzziness so as not to be too accurate in identifying specific locations? Either way, the map locates crimes at points that don't actually locate crimes and probably fails to show us the true pattern. It's a dangerous story to tell. It's also laziness...the data are most likely reported to a nearest enumeration point and that defines the map. No attempt to process it into something more meaningful.

As readers of Treasure Island know full well, X didn't actually mark the spot where the treasure was buried...and this national map of crime perpetuates this myth by showing inaccurate locations. I look forward to version 3 in 2016 when they might actually address the data and cartographic issues of the map rather than just the cosmetic.The KISS principle used to mean "Keep It Simple Stupid" but now seems to be more appropriately "Keep It Slick Stupid".  UI/UX is not the same as cartographic design.



2 comments: