Chris Osborne posted a comment to an earlier entry drawing our attention to a MapTube project mapping crime in London. This deserves a fuller article as it demonstrates a number of interesting points.
1. MapTube itself is worth a mention. This initiative from UCL’s Centre for Advanced Spatial Analysis (CASA) is a great demonstrator of the power of map-based information. There is a mass of interesting information and tools behind MapTube with the GeoVue project of particular interest. The GMapCreator software that has been created as part of this is useful for Google Maps mashers.
2. Privacy has come up as an issue for the Power of Information agenda. We have set ourselves the task of only dealing with data that is not personal. In most cases the distinction is clear, but there are areas such as crime mapping where there is some uncertainty about whether publishing detailed information would have implications in terms of data protection law. The point to consider is the extent to which individuals, especially victims, can be identified when details of individual crime incidents are shown. This quote from the Data Protection Act 1998 tells us when data has to be regarded as personal:
“personal data” means data which relate to a living individual who can be identified—
(a) from those data, or
(b) from those data and other information which is in the possession of, or is likely to come into the possession of, the data controller,
The issue for some uses of apparently anonymous public data is in this second clause, i.e. does it relate to a real person when the data is combined with other information. The quantity of data now available online and the very power of Web 2.0 technologies makes it increasingly likely that data can be linked up to identify an individual.
The MapTube crime map links through pins to published newspaper articles which name individual crime victims. In the cases shown on the map, the victims are tragically deceased and therefore outside the specific definition of personal data, but this does provide a graphic illustration of the general phenomenon of linkage of crimes to victims that is worrying the regulators.
Richard Allan, Task Force Chair
Power of Information Task Force 
Steve Feldman has made an excellent point at http://feeds.feedburner.com/~r/Giscussions/~3/343302081/pinpointing-crme.html – that some (full) postcodes identify a place much more precisely than others. (Typically a full postcode identifies about 15 properties, but it may be fewer, or more.)
It’s certainly one to chew over; perhaps the solution might be that crimes are mapped by semi-full postcode – eg SW1 3E-, leaving off the final identifier.
In the case of murders, though, that does cross over to a situation where it’s not “personal” information any more. But… what if you had a murder and a wounding in the same place?
Complicated indeed.
That’s the catch with freedom of data, though. Your quote from the DPA talks about ‘information which is in the possession of, or is likely to come into the possession of’. It doesn’t say ‘could potentially come into the possession of’.
The more data gets put out there, officially or otherwise, the more could potentially be accessed by the data controller. The word ‘likely’ seems to come from an age when things had to be provided, person to person, on floppy disk. Those days are long gone.
I’m looking forward to someone ruling on whether crime mapping is actually feasible in the current legal context.
(PS: The BBC’s Mark Easton made an interesting point re murder data earlier in the week; it’s the only figure where you can be pretty certain the crime will be reported. Any other measure faces allegations of under-reporting. Fair point, I thought.)
There is significant breathing room when it comes to releasing crime data.
1. There is no need for publicly released data to be high resolution; knowing that the crime occurred on a street or neighbourhood is more than adequate for public knowledge.
2. Linking an individual crime to an individual will be extremely difficult unless it is a significant crime that attracts media attention. Everybody knows if there is a murder down the street, however most minor crimes are unknown unless your neighbours chose to share the information with you.
3. Having worked with police data I can tell you that there are no common standards for recording location. Geocoding is sketchy at best.
4. Everyblock.org have been displaying US crime data by block resolution for years without complaint.
The issues of geocoding like postcodes being too granular and revealing identifiable information is not new, will not go away…but there are ways around it. While I was at University I picked up some extra cash developing a sales analysis tool for a pharmaceutical company. They knew what/where they spent money on sales and marketing, and were provided with information back on sales of their products. However, because of ethical rules they were not allowed to know the sales from an individual pharmacy. A postcode segment could include 10 pharmacies or just one, so they came up with the idea of ‘bricks’, which were units that included one or more postcode. That meant you could mask it where it was too granular, without making it stupidly blunt in more dense urban areas. Algorithmically, this would not be hard to do for any kind of geocode data. For example on crime data, you could set a household/business per postcode threshold, and if it fell below that it was combined with an adjacent one. Then when doing a search you match the postcode agains the ‘brick’ (or whatever) and return data based on that. As long as you can define an unacceptable level of granularity based on available and processable data, you could design this up front into any data interface quite easily. Retrofitting it would be a blighter though!
Pingback: London Crime Mapping Live « Power of Information Task Force