Crime sensing with big data
A fascinating recent British Journal of Criminology article examined the usefulness of Twitter in predicting crime rates in London boroughs.
Crime sensing with big data: the affordances and limitations of using open-source communications to estimate crime patterns was written by Matthew Williams, Pete Burnap and Luke Sloan.
The paper explored the hypothesis that disorder-related posts on Twitter are associated with actual police crime rates.
The hypothesis has its routes in the “broken windows” theory in criminology which, in simple terms, is that visible signs of neighbourhood degeneration are causally linked to crime rates.
The researchers looked at three data sources to test their hypothesis:
- Met Police crime data for nine crime categories in 28 London boroughs over a 12-month
- The UK Census 2011
- UK tweets accessed via the the Twitter streaming Application Programming Interface using the Cardiff Online Social Media Observatory (COSMOS) software platform; once this dataset was reduced to those geolocated in the 28 London boroughs over this 12 month period, it consisted of 8,417,438 tweets.
The researchers developed a number of “broken windows” text indicators from previous research on policing priorities and identified key terms to search for in the tweets. To illustrate this, the researchers provided some example tweets:
‘New graffiti at the end of my street. How did they reach that high!?’
‘Community allotment was vandalized today. Why would someone do this?
‘More illegal dumping in Shoreditch. When will @hackneycouncil sort this out?!’
‘RT if you think we should use discarded card receipts to identify litterers!’
The researchers found that Tweet frequency was particularly associated with four of the nine crime types studied:
- Burglary of a dwelling
- Criminal damage
- Violence against the person
A particularly interesting finding was that tweets containing mentions of “broken windows” indicators are positively correlated with criminal damage, theft from a motor vehicle, possession of drugs and violence — but only in low crime areas. The same tweets were negatively correlated with burglary of a dwelling, commercial burglary and theft of a motor vehicle in high crime areas. The researchers speculate that this is because crimes of this nature are unremarkable in high crime areas and therefore do not stimulate the same anger and motivation to broadcast their occurrence as they do in low crime areas.
The study and use of big data, especially social media big data, are still at relatively early stages and there are a host of methodological problems to tackle which can often be particularly tricky because patterns and culture of social media use are changing so rapidly.
Nevertheless, this is an intriguing study whose authors conclude:
The models provide some preliminary, but nevertheless, encouraging results that indicate open-source communications, in particular from Twitter, have potential for measuring the breakdown of social and physical order at the borough level.
I have already mentioned the limitation that this approach (especially in terms of the broken windows crimes) seems to work better in low-crime areas. However, the opportunity of this sort of approach is that it does tend to capture the views of young men, who are often under-represented in conventional surveys.
The use of social media in predicting real-time disorder is a growing science in the US; it is pleasing to report on innovative research conducted in the UK for a change.
Watch this space to see the latest developments.
All innovation posts are kindly sponsored by Socrates 360 which provides a complete solution for staff, prisoners, probationers, etc. combining engaging content, simple set-up and an easy tracking system. Socrates 360 has no influence over editorial content.