Page 50 - BBC Focus - August 2017
P. 50

PREDIC TIVE





      POLICING                                                         these locations?” he says.


                                                                       “Instead, we should be asking
                                                                       why this location has a persistent
                                                                       crime problem, and what we can
                                                                       do to keep it from happening.”
       CAN WE PREDICT CRIMES BEFORE THEY HAPPEN?                       Eck would prefer it if the police
                                                                       encouraged owners of
                                                                       businesses and other properties
                                                                       highlighted as crime hotspots to
                                                                       step in and make changes, such
                It’s 4:30am on a Friday morning   on those that help to predict   as shops with high shoplifting
                in August and there’s a heavy   when and where the next   rates repositioning displays.
                police presence in a quiet London   earthquake aftershock will be, or   Critics have raised other concerns
                suburb. It’s a respectable, leafy   how a disease will spread.  too, such as the possibility of
                area and right now, nothing is   These algorithms generate   crimes simply shifting to other
                happening. In fact, it’s been quiet   information that police officers   locations when problem areas
                for the past few days. But the   can act on, and it seems to work.   are targeted by the police.
                officers are on high alert. They’ve   In tests, their predictive powers   But predictive policing is
                been sent at the say-so of a   appear to outperform the more   becoming more and more
                computer that’s calculated, on   traditional techniques used by   widespread, and it could be
                the basis of the data fed into it,   crime analysts. Their successes   about to change radically. Earlier
                that a wave of break-ins is highly   have led to predictive policing   this year, a bunch of
                likely within the next 24 hours. In   being adopted by several US   mathematicians led by Prof Mark
                other words, they’re policing   police departments, such as   Girolami at Imperial College
                crimes that they think will   California and Arizona, as well as   London were awarded £3m from
                happen, rather than ones that   Kent Police in the UK.  the government to take
                have happened. This is predictive   But not everyone’s convinced   predictive policing to the next
                policing. And it’s about to get   about predictive policing – or   level. Whereas today’s tools just
                much, much more sophisticated.  how it’s implemented at least.   rely on crime data – such as the
                  The idea of predicting where   Among them is criminologist Prof   locations, dates and times of
                crimes will take place isn’t new.   John Eck at the University of   incidents – Girolami and his team
                For decades now, police forces in   Cincinnati. His problem isn’t so   will be working on how to
                the UK and US have been creating   much with the predictive policing   integrate the likes of Twitter
                ‘hotspot’ maps that identify the   software itself, but the idea of   feeds, newspaper reports and
                areas where most incidents are   sending out large numbers of   socioeconomic data to sharpen
                taking place, and then sending   staff to patrol problems   the predictions. Text documents
                more police officers to those   highlighted by the algorithms.   will be converted, or ‘coded’, into
                areas. Predictive policing takes   “Why would you want to keep   numerical representations, with
                this to the next level, crunching   sending large amounts of   counts of words and phrases
                big data using algorithms based   expensive public servants to   – such as descriptions of assaults
                                                                       or break-ins – to highlight
                                                                       geographical areas of concern.
                                                            Staff from the   “All of these streams of
                                                            University of
                                                              California   information will be coded and
                                                            demonstrate   integrated using our ‘secret
                                                              predictive   sauce’,” says Girolami, referring to
                                                            policing with
                                                             Los Angeles   the complex maths that will draw
                                                                police  all of this disparate data together.
                                                                         What’s more, this new
                                                                       predictive tool aims to work out
                                                                       the extent to which crime will be   PHOTO:  SHUTTERSTOCK  ILLUSTRATION: VLADO KRIZAN
                                                                       displaced to a neighbouring area
                                                                       when the number of police in the
                                                                       original area suddenly shoots up.
                                                                       “Our models will be able to
                                                                       propagate what would happen,”
                                                                       says Girolami.


       50
   45   46   47   48   49   50   51   52   53   54   55