00:00 At the heart of predictive policing lies a big promise, to stop crime before it is even
00:06 committed.
00:07 This is done by combing through masses of data.
00:10 Some programs focus on potential crime scenes.
00:13 If they find links between previous incidents, say that they all occurred at the same time
00:17 and place, or even during the same weather conditions, they recommend that police zero
00:21 in on those locations.
00:25 Other programs predict what they consider to be potential offenders.
00:29 They scan the criminal and personal histories of individuals for so-called risk factors,
00:34 like previous arrests, or even dropping out of school, and come up with a list of who
00:38 is likely to break the law.
00:45 Law enforcement agencies deploy these programs in over 50 countries, Western democracies,
00:51 and authoritarian governments.
00:58 Opponents of predictive policing say it makes cities safer, but the approach has caused
01:03 outrage among human rights advocates.
01:07 They warn that it increases discrimination.
01:11 The programs, for instance, often flag low-income communities or minority neighborhoods as alleged
01:17 hot spots, prompting the police to patrol those areas more than others.
01:23 This in turn generates even more data and sets off a vicious cycle of discrimination,
01:28 flagging these areas over and over again.
01:32 Similarly, the programs tend to single out low-income people and minorities as potential
01:37 offenders.
01:38 Existing police data, whether in New Delhi, whether in the US, whether in Germany, whether
01:44 in Australia, embed a lot of historical discrimination.
01:48 They do embed systemic problems such as racism or casteism or even sexism.
01:54 Now we're taking all of this data that is seemingly objective and correct, but is in
02:00 reality quite biased and discriminatory, especially towards minority populations.
02:05 We take that as being the ground truth on which the computer learns what patterns exist.
02:12 Public awareness for the risks of predictive policing and biases against some social groups
02:16 is growing.
02:18 In the US, amid a debate over systemic racism and how to reform policing, some police departments
02:25 stopped using facial recognition and predictive policing.
02:29 But in other countries, the technology is still in demand.
02:33 In countries that don't have strong data protection legislation or where governments are more
02:38 open to experimenting with technologies on people, we see that predictive policing is
Comments