WEBVTT

1
00:00:00.000 --> 00:00:01.480
[police radio]

2
00:00:01.480 --> 00:00:02.630
<v Issie>Police in Los Angeles</v>

3
00:00:02.630 --> 00:00:04.700
are trying to predict the future.

4
00:00:04.700 --> 00:00:06.500
<v ->We know where crime happened yesterday.</v>

5
00:00:06.500 --> 00:00:09.440
But where is it going to happen tomorrow and the next day?

6
00:00:09.440 --> 00:00:10.750
<v Issie>And they're not alone.</v>

7
00:00:10.750 --> 00:00:13.460
More and more departments are using data driven algorithms

8
00:00:13.460 --> 00:00:15.260
to forecast crime.

9
00:00:15.260 --> 00:00:16.630
<v ->It's all about improving accuracy,</v>

10
00:00:16.630 --> 00:00:18.750
and improving our effectiveness and our efficiency,

11
00:00:18.750 --> 00:00:20.530
and our service to the community.

12
00:00:20.530 --> 00:00:21.810
<v ->When you see more police officers.</v>

13
00:00:21.810 --> 00:00:23.208
You see the lights, you hear the sirens.

14
00:00:23.208 --> 00:00:25.030
The high visibility of officers

15
00:00:25.030 --> 00:00:27.370
does deter crime in certain areas.

16
00:00:27.370 --> 00:00:30.030
Proponents say it's helping to reduce crime.

17
00:00:30.030 --> 00:00:32.630
But other wonder, at what cost?

18
00:00:32.630 --> 00:00:35.160
<v ->I think it targets and it just justifies</v>

19
00:00:35.160 --> 00:00:35.993
in their eyes

20
00:00:35.993 --> 00:00:37.310
racial profiling.

21
00:00:37.310 --> 00:00:39.410
<v ->I just really worry about the world</v>

22
00:00:39.410 --> 00:00:41.060
that my son is gonna enter into,

23
00:00:41.060 --> 00:00:42.360
and become a young man in.

24
00:00:42.360 --> 00:00:45.440
I don't want people to look at him and prejudge him

25
00:00:45.440 --> 00:00:46.840
that could be life or death

26
00:00:46.840 --> 00:00:48.470
if someone calls the police on you.

27
00:00:48.470 --> 00:00:50.000
<v ->The algorithm is always going to augment</v>

28
00:00:50.000 --> 00:00:50.930
the system that it's in.

29
00:00:50.930 --> 00:00:53.250
And if the system itself is biased,

30
00:00:53.250 --> 00:00:54.490
is unjust,

31
00:00:54.490 --> 00:00:56.910
then the algorithm's gonna replicate that.

32
00:00:56.910 --> 00:00:59.370
<v ->Predictive policing is changing policing</v>

33
00:00:59.370 --> 00:01:00.760
in your city now.

34
00:01:00.760 --> 00:01:02.920
And when you start asking the hard questions about.

35
00:01:02.920 --> 00:01:04.420
How does it impact civil rights?

36
00:01:04.420 --> 00:01:05.830
How does it impact privacy?

37
00:01:05.830 --> 00:01:07.140
How does it impact suspects?

38
00:01:07.140 --> 00:01:09.340
How does it impact officers doing their job?

39
00:01:14.590 --> 00:01:16.240
<v ->Alright here's how where gonna work for tonight.</v>

40
00:01:16.240 --> 00:01:18.120
Sergeant Flores will be out as 160.

41
00:01:18.120 --> 00:01:19.240
<v Issie>Say you're living in Los Angeles.</v>

42
00:01:19.240 --> 00:01:21.150
And you just got out of prison.

43
00:01:21.150 --> 00:01:23.340
The LAPD could be keeping tabs on you

44
00:01:23.340 --> 00:01:26.210
through a program known as Operation LASER.

45
00:01:26.210 --> 00:01:28.410
It uses crime, arrest and field data

46
00:01:28.410 --> 00:01:31.210
to determine where violent crimes are likely to take place.

47
00:01:31.210 --> 00:01:32.910
And who will perpetrate them.

48
00:01:32.910 --> 00:01:35.290
<v ->LASER predicts who will commit a crime</v>

49
00:01:35.290 --> 00:01:37.400
before crime actually happens.

50
00:01:37.400 --> 00:01:39.740
<v ->Groups like the Stop LAPD Spying Coalition</v>

51
00:01:39.740 --> 00:01:42.030
are sounding the alarm in local communities.

52
00:01:42.030 --> 00:01:44.500
They say predictive policing strategies like LASER

53
00:01:44.500 --> 00:01:46.780
Disproportionately target low income people

54
00:01:46.780 --> 00:01:48.920
and communities of color.

55
00:01:48.920 --> 00:01:51.230 line:15% 
<v ->Why do you believe that this data is subjective?</v>

56
00:01:51.230 --> 00:01:53.890 line:15% 
<v ->The data is subjective and the data is biased</v>

57
00:01:53.890 --> 00:01:57.180 line:15% 
because policing practices have always impacted

58
00:01:57.180 --> 00:01:58.460 line:15% 
black and brown

59
00:01:58.460 --> 00:01:59.293
and poor people.

60
00:01:59.293 --> 00:02:01.100
Now we have predictive policing

61
00:02:01.100 --> 00:02:02.720
coming in with a veneer of science

62
00:02:02.720 --> 00:02:03.630
trying to tell us

63
00:02:03.630 --> 00:02:06.710
that somehow making policing more objective.

64
00:02:06.710 --> 00:02:07.610
But what we're finding is

65
00:02:07.610 --> 00:02:10.800
that it's only augmenting the way that policing operates.

66
00:02:10.800 --> 00:02:12.440
The LASER program tries to predict

67
00:02:12.440 --> 00:02:15.020
who might commit violent crime using a point system.

68
00:02:15.020 --> 00:02:16.000 line:15% 
For example,

69
00:02:16.000 --> 00:02:18.495 line:15% 
If you're on probation that's five points on your record.

70
00:02:18.495 --> 00:02:20.719
If you're in a gang that's another five.

71
00:02:20.719 --> 00:02:23.490
A stop by police can get you another point.

72
00:02:23.490 --> 00:02:24.830
The people with the highest scores

73
00:02:24.830 --> 00:02:26.470
in each of LA's 21 divisions

74
00:02:26.470 --> 00:02:29.740
are placed on a list called the Chronic Offender Bulletin.

75
00:02:29.740 --> 00:02:31.990 line:15% 
<v ->You tally up all those numbers.</v>

76
00:02:31.990 --> 00:02:34.500 line:15% 
Some folks end up with

77
00:02:34.500 --> 00:02:35.333 line:15% 
two points.

78
00:02:35.333 --> 00:02:38.240 line:15% 
Some folks end up with 40 points.

79
00:02:38.240 --> 00:02:40.670
Those that end up with the highest number of points,

80
00:02:40.670 --> 00:02:41.550
we'll take a look and see.

81
00:02:41.550 --> 00:02:42.570
Where are they now?

82
00:02:42.570 --> 00:02:44.270
<v ->The most troubling thing</v>

83
00:02:44.270 --> 00:02:45.510
about the Chronic Offender Bulletin,

84
00:02:45.510 --> 00:02:48.540
it' re-criminalizing people who have already been on parole.

85
00:02:48.540 --> 00:02:50.630
Who have served their time.

86
00:02:50.630 --> 00:02:53.170
<v ->It seems kind of hard to avoid police contact</v>

87
00:02:53.170 --> 00:02:56.100
when that's your home and this is your history.

88
00:02:56.100 --> 00:02:58.210
And the contact is coming to you.

89
00:02:58.210 --> 00:02:59.767
<v ->Well not necessarily.</v>

90
00:02:59.767 --> 00:03:01.750
If you live in the area,

91
00:03:01.750 --> 00:03:02.950
and you did time,

92
00:03:02.950 --> 00:03:04.070
and you've got a job,

93
00:03:04.070 --> 00:03:06.200
and you're doing everything you're supposed to be doing,

94
00:03:06.200 --> 00:03:07.250
then you're not gonna be getting

95
00:03:07.250 --> 00:03:08.730
that constant police contact.

96
00:03:08.730 --> 00:03:10.920
<v Issie>Police say they're just being proactive.</v>

97
00:03:10.920 --> 00:03:12.870
<v ->What we don't want is for that person</v>

98
00:03:12.870 --> 00:03:14.977
to recommit another crime.

99
00:03:14.977 --> 00:03:17.580
Go back to jail and keep on the cycles.

100
00:03:17.580 --> 00:03:18.960
We try to get ahead of it.

101
00:03:18.960 --> 00:03:20.660
<v Issie>The problem is the public has no way</v>

102
00:03:20.660 --> 00:03:22.190
of knowing who's being targeted,

103
00:03:22.190 --> 00:03:24.610
which is why the Stop LAPD Spying Coalition

104
00:03:24.610 --> 00:03:27.610
is suing the city for more information.

105
00:03:27.610 --> 00:03:29.720
Police are supposed to inform individuals

106
00:03:29.720 --> 00:03:30.890
if they're on the list.

107
00:03:30.890 --> 00:03:31.730
But we reached out

108
00:03:31.730 --> 00:03:34.466
to the ACLU, Legal Aid and several defense attorneys

109
00:03:34.466 --> 00:03:36.240
and none of them knew of anyone

110
00:03:36.240 --> 00:03:38.550
who had been contacted by LAPD.

111
00:03:38.550 --> 00:03:39.920
<v ->We feel that the people have the right</v>

112
00:03:39.920 --> 00:03:42.600
to know what is going on in this program.

113
00:03:42.600 --> 00:03:44.620
They may not know the name of LASER.

114
00:03:44.620 --> 00:03:47.220
And they may not know the specifics of how it operates.

115
00:03:47.220 --> 00:03:49.960
But I think they do know exactly what's happening

116
00:03:49.960 --> 00:03:50.940
that they're being monitored.

117
00:03:50.940 --> 00:03:52.880
Cause people in the community they speak about it.

118
00:03:52.880 --> 00:03:55.210
They say, the officer came up to me and said,

119
00:03:55.210 --> 00:03:57.190
I see you, I know your brother.

120
00:03:57.190 --> 00:03:59.810
I got your brother, I'm gonna get you.

121
00:03:59.810 --> 00:04:02.010 line:15% 
<v ->It is one part social control.</v>

122
00:04:02.010 --> 00:04:03.150 line:15% 
Hey we're watching you.

123
00:04:03.150 --> 00:04:04.820 line:15% 
And it's one part information clashing.

124
00:04:04.820 --> 00:04:06.445 line:15% 
We are seeing who you are associating with.

125
00:04:06.445 --> 00:04:08.450 line:15% 
The car you're driving.

126
00:04:08.450 --> 00:04:09.830
Who you're hanging out with in your neighborhood.

127
00:04:09.830 --> 00:04:11.230
And they're recording all that,

128
00:04:11.230 --> 00:04:15.080
and putting it back into a larger, ever growing database

129
00:04:15.080 --> 00:04:17.490
that's used for investigative purposes.

130
00:04:17.490 --> 00:04:18.520
<v Issie>And LASER is just one</v>

131
00:04:18.520 --> 00:04:20.810
of several new smart policing strategies.

132
00:04:20.810 --> 00:04:23.430
Some departments are also considering drone surveillance,

133
00:04:23.430 --> 00:04:24.720
and body cameras equipped

134
00:04:24.720 --> 00:04:27.000
with facial recognition technology.

135
00:04:27.000 --> 00:04:29.690
This video from the maker of an AI powered body cam

136
00:04:29.690 --> 00:04:31.760
demonstrates how the technology could be used

137
00:04:31.760 --> 00:04:33.410
to find a missing child.

138
00:04:33.410 --> 00:04:34.243
Not shown,

139
00:04:34.243 --> 00:04:36.680
how they could use it to track suspects.

140
00:04:36.680 --> 00:04:38.610
Meanwhile, cities across the country

141
00:04:38.610 --> 00:04:40.460
have embraced a predictive policing approach

142
00:04:40.460 --> 00:04:42.450
that focuses not on high risk people

143
00:04:42.450 --> 00:04:44.370
but on high risk places.

144
00:04:44.370 --> 00:04:46.720
Putting entire neighborhoods under watch.

145
00:04:46.720 --> 00:04:48.198
Critics worry it's targeting areas

146
00:04:48.198 --> 00:04:50.920
that are already over policed.

147
00:04:50.920 --> 00:04:53.130
<v ->The problem is that if you're already policing</v>

148
00:04:53.130 --> 00:04:54.300
in a certain neighborhood,

149
00:04:54.300 --> 00:04:56.570
you're gonna be predisposed to seeing data

150
00:04:56.570 --> 00:04:57.570
that's now gonna tell you

151
00:04:57.570 --> 00:05:00.280
to keep putting more and more officers in those areas.

152
00:05:00.280 --> 00:05:01.700
<v Issie>Bari Williams is an attorney</v>

153
00:05:01.700 --> 00:05:03.860
who advises startups in the Bay Area.

154
00:05:03.860 --> 00:05:05.490
<v ->I absolutely love technology.</v>

155
00:05:05.490 --> 00:05:06.860
I work in the sector.

156
00:05:06.860 --> 00:05:09.330
But I think that we need to be mindful

157
00:05:09.330 --> 00:05:11.380
of how technology is created.

158
00:05:11.380 --> 00:05:12.740
And how it's used.

159
00:05:12.740 --> 00:05:15.670
And how it can be used in various purposes.

160
00:05:15.670 --> 00:05:17.060
<v Issie>One popular tool focused</v>

161
00:05:17.060 --> 00:05:19.100
on where crime will happen is PredPol.

162
00:05:19.100 --> 00:05:21.590
Today more than 60 departments nationwide

163
00:05:21.590 --> 00:05:22.860
use it to forecast crimes

164
00:05:22.860 --> 00:05:24.697
like burglaries and car break ins.

165
00:05:24.697 --> 00:05:26.377
<v ->Everybody has their predictive policing already.</v>

166
00:05:26.377 --> 00:05:28.320
<v Issie>In the LAPD's foothill division,</v>

167
00:05:28.320 --> 00:05:29.880
one of the first to use PredPol,

168
00:05:29.880 --> 00:05:31.690
it starts with the daily roll call.

169
00:05:31.690 --> 00:05:34.410
Officers get a map labeled with 10 to 20 hotspots

170
00:05:34.410 --> 00:05:36.590
they're encouraged to visit during their shifts.

171
00:05:36.590 --> 00:05:38.360
The more time they spend in those areas,

172
00:05:38.360 --> 00:05:40.400
the more likely they are to deter crime.

173
00:05:40.400 --> 00:05:41.703
Or so the thinking goes.

174
00:05:41.703 --> 00:05:42.760
<v Officer>Alright, let's go to work.</v>

175
00:05:42.760 --> 00:05:45.380
<v ->This is today's predictive policing boxes.</v>

176
00:05:45.380 --> 00:05:47.610 line:15% 
We're gonna cover the three boxes that are in my area.

177
00:05:47.610 --> 00:05:49.964
In the north end of Pocoima.

178
00:05:49.964 --> 00:05:51.260
<v Issie>PredPol forecasts crimes</v>

179
00:05:51.260 --> 00:05:53.770
based on patterns from the last several years.

180
00:05:53.770 --> 00:05:56.490
It analyzes three elements of crime data,

181
00:05:56.490 --> 00:05:59.250
the crime type, the location and the time.

182
00:05:59.250 --> 00:06:00.930
It then runs them through an algorithm

183
00:06:00.930 --> 00:06:04.410
and spits out 500 by 500 foot hotspots, or boxes

184
00:06:04.410 --> 00:06:06.870
that officers should keep an eye on.

185
00:06:06.870 --> 00:06:09.880
<v ->Right now we're experiencing a lot of property crime.</v>

186
00:06:09.880 --> 00:06:11.610
That's what our predictive policing boxes

187
00:06:11.610 --> 00:06:13.300
are gonna focus on.

188
00:06:13.300 --> 00:06:14.780
<v Issie>Officer Steve Nunez has worked</v>

189
00:06:14.780 --> 00:06:17.140
in the FootHill Division for 16 years.

190
00:06:17.140 --> 00:06:19.820
What are you looking for as we cruise around?

191
00:06:19.820 --> 00:06:21.450
<v ->Looking for anything suspicious.</v>

192
00:06:21.450 --> 00:06:23.220
Even sometimes looking for cars

193
00:06:23.220 --> 00:06:25.530
that look like they've been sitting here for awhile.

194
00:06:25.530 --> 00:06:28.280
Anyone that is maybe trying to open car doors.

195
00:06:28.280 --> 00:06:30.210
Just anything that doesn't fit.

196
00:06:30.210 --> 00:06:31.520
At first we were kind of,

197
00:06:31.520 --> 00:06:34.160
I don't know if we're gonna follow this predictive policing.

198
00:06:34.160 --> 00:06:36.100
Having a computer tell us where to go.

199
00:06:36.100 --> 00:06:38.370
<v Issie>But he's become a believer in PredPol.</v>

200
00:06:38.370 --> 00:06:40.460
<v ->Crime has gone down dramatically.</v>

201
00:06:40.460 --> 00:06:42.240
Part of it I believe is predictive policing.

202
00:06:42.240 --> 00:06:45.220
Part of it's just the community engagement.

203
00:06:45.220 --> 00:06:46.770
Has to do with the officers

204
00:06:46.770 --> 00:06:49.040
going out there and being proactive.

205
00:06:49.040 --> 00:06:50.540
Don't leave it out here for a month

206
00:06:50.540 --> 00:06:51.570
and all that stuff like that.

207
00:06:51.570 --> 00:06:52.950
Like I said, I won't impound it,

208
00:06:52.950 --> 00:06:53.920
cause I know it's his.

209
00:06:53.920 --> 00:06:55.240
So I'll let you guys deal with it.

210
00:06:55.240 --> 00:06:56.840
But at least get it off the street.

211
00:06:56.840 --> 00:06:57.920
<v ->Alright, man.</v>
<v ->Cool, man.</v>

212
00:06:57.920 --> 00:06:58.960
You guys be good, alright?

213
00:06:58.960 --> 00:07:00.190
<v ->You too.</v>

214
00:07:00.190 --> 00:07:01.630
<v Issie>The fundamental question here is,</v>

215
00:07:01.630 --> 00:07:03.360
can we trust the data and algorithms

216
00:07:03.360 --> 00:07:06.240
that police are using to predict crimes?

217
00:07:06.240 --> 00:07:08.720 line:15% 
<v ->Algorithms and big data</v>

218
00:07:08.720 --> 00:07:10.470 line:15% 
simply take inputs,

219
00:07:10.470 --> 00:07:12.456
crunch 'em up to create outputs.

220
00:07:12.456 --> 00:07:14.200
If your inputs are biased,

221
00:07:14.200 --> 00:07:15.850
your outputs are gonna be biased.

222
00:07:15.850 --> 00:07:16.870
<v Issie>You don't need to look far</v>

223
00:07:16.870 --> 00:07:19.360
to find troubling examples of algorithmic bias.

224
00:07:19.360 --> 00:07:21.530
In 2016, Microsoft introduced

225
00:07:21.530 --> 00:07:23.490
an AI chat bot called Tay,

226
00:07:23.490 --> 00:07:24.940
which quickly started repeating

227
00:07:24.940 --> 00:07:26.452
racist and anti-Semitic comments

228
00:07:26.452 --> 00:07:28.440
it picked up on Twitter.

229
00:07:28.440 --> 00:07:30.790
Then there were those problematic Google auto-completes,

230
00:07:30.790 --> 00:07:33.060
which are based on popular human searches.

231
00:07:33.060 --> 00:07:35.840
In January, a Google search for "islamists are"

232
00:07:35.840 --> 00:07:38.510
turned up results like, islamists are evil.

233
00:07:38.510 --> 00:07:41.550
And "hitler is" turned up, hitler is my hero.

234
00:07:41.550 --> 00:07:42.870
And there was also the time,

235
00:07:42.870 --> 00:07:44.200
the Google Images algorithm

236
00:07:44.200 --> 00:07:46.696
mistook black people for gorillas.

237
00:07:46.696 --> 00:07:48.620
What happens if predjudiced data

238
00:07:48.620 --> 00:07:49.883
makes it's way into police work?

239
00:07:49.883 --> 00:07:50.930
<v ->It's a seven.</v>

240
00:07:50.930 --> 00:07:53.380
<v Issie>For Bari Williams that risk is personal.</v>

241
00:07:53.380 --> 00:07:54.330
<v ->The biggest fear I have</v>

242
00:07:54.330 --> 00:07:55.930
with predictive policing and my son

243
00:07:55.930 --> 00:07:58.140
is that he will be profiled.

244
00:07:58.140 --> 00:08:00.550
I'm sure he probably will experience that

245
00:08:00.550 --> 00:08:02.280
at some point in his life already.

246
00:08:02.280 --> 00:08:04.960
But I don't need something making that easier.

247
00:08:04.960 --> 00:08:05.793
<v ->Did you have a good day?</v>

248
00:08:05.793 --> 00:08:06.626
<v ->Yeah.</v>

249
00:08:06.626 --> 00:08:07.470
<v Issie>Williams lives in Oakland,</v>

250
00:08:07.470 --> 00:08:10.420
a city that so far has decided not to use PredPol.

251
00:08:10.420 --> 00:08:11.253
Still she says,

252
00:08:11.253 --> 00:08:13.670
the image of black children being fatally misslabled

253
00:08:13.670 --> 00:08:16.970
as a threat by police looms large in her mind.

254
00:08:16.970 --> 00:08:18.650
<v ->In the case of Tamir Rice,</v>

255
00:08:18.650 --> 00:08:20.820
you have a child who was playing with a BB gun.

256
00:08:20.820 --> 00:08:22.890
And the police just rolled up.

257
00:08:22.890 --> 00:08:24.683
And within three seconds shot him.

258
00:08:25.530 --> 00:08:27.430
That is a real concern for me.

259
00:08:27.430 --> 00:08:30.230
My son and I like to go certain places, and

260
00:08:30.230 --> 00:08:31.850
if the neighborhoods are not awesome,

261
00:08:31.850 --> 00:08:33.700
and predictive policing is telling you

262
00:08:33.700 --> 00:08:36.720
to send more police into that area,

263
00:08:36.720 --> 00:08:37.920
and we're there.

264
00:08:37.920 --> 00:08:41.000
I'm not gonna allow my child to go play in the street.

265
00:08:41.000 --> 00:08:43.480
I'm not gonna allow my child to run in front of me.

266
00:08:43.480 --> 00:08:46.860
I don't want anything to give the police

267
00:08:46.860 --> 00:08:49.690
a false sense that he could've been

268
00:08:49.690 --> 00:08:51.530
engaged in something he wasn't.

269
00:08:51.530 --> 00:08:53.900
<v ->The activists and the police</v>

270
00:08:53.900 --> 00:08:57.370
are all coming from the point that makes sense for them.

271
00:08:57.370 --> 00:08:58.840
I think from a community perspective,

272
00:08:58.840 --> 00:09:00.390
that feels over policed.

273
00:09:00.390 --> 00:09:03.216
That feels like there is racial injustice in policing.

274
00:09:03.216 --> 00:09:07.110
It's wise to be concerned about this new technology

275
00:09:07.110 --> 00:09:09.580
that's going to give police more power.

276
00:09:09.580 --> 00:09:11.100
If you're a police officer,

277
00:09:11.100 --> 00:09:13.790
you understand you have an incredibly difficult job.

278
00:09:13.790 --> 00:09:15.580
And adding new technologies might help.

279
00:09:15.580 --> 00:09:17.830
You kinda wanna get the benefit of the doubt.

280
00:09:17.830 --> 00:09:20.140
<v ->We've had crime rises over the last few years.</v>

281
00:09:20.140 --> 00:09:22.340
We've had crime reduction.

282
00:09:22.340 --> 00:09:23.610
It is a tool.

283
00:09:23.610 --> 00:09:25.310
It's not a panacea for crime.

284
00:09:25.310 --> 00:09:26.920
That's what a lot of people believe.

285
00:09:26.920 --> 00:09:28.570
They believe that you utilize PredPol,

286
00:09:28.570 --> 00:09:30.080
crime's gonna go away.

287
00:09:30.080 --> 00:09:30.920
False.

288
00:09:30.920 --> 00:09:31.753
That is a farce.

289
00:09:31.753 --> 00:09:32.722
That is not gonna happen.

290
00:09:32.722 --> 00:09:34.250
It's an indicator.

291
00:09:34.250 --> 00:09:35.400
It takes you there.

292
00:09:35.400 --> 00:09:36.280
It could happen.

293
00:09:36.280 --> 00:09:37.113
It could not.

294
00:09:37.113 --> 00:09:38.080
It's a probability.

295
00:09:38.080 --> 00:09:41.110
<v ->The question about whether predictive policing works</v>

296
00:09:41.110 --> 00:09:42.370
is still open.

297
00:09:42.370 --> 00:09:44.280
<v Issie>One study found that PredPol's algorithms</v>

298
00:09:44.280 --> 00:09:47.890
correctly predicted crime just 4.7% of the time.

299
00:09:47.890 --> 00:09:49.740
Still that's higher than the human analysts

300
00:09:49.740 --> 00:09:52.710
who got it right just 2.1% of the time.

301
00:09:52.710 --> 00:09:53.570
It's worth noting that

302
00:09:53.570 --> 00:09:56.150
the study was done by the company's co-founder.

303
00:09:56.150 --> 00:09:57.900
And then there's the question of bias.

304
00:09:57.900 --> 00:09:59.360
<v ->You start coming into the Pacoima area,</v>

305
00:09:59.360 --> 00:10:01.390
which is south of the 210 freeway.

306
00:10:01.390 --> 00:10:03.710
Now you're dealing with predominantly Hispanic families.

307
00:10:03.710 --> 00:10:05.400
A lot of immigrants in that area.

308
00:10:05.400 --> 00:10:08.510
<v ->And I notice there are more boxes there.</v>

309
00:10:08.510 --> 00:10:09.670
The day we visited most

310
00:10:09.670 --> 00:10:11.590
of the PredPol boxes in the Foothill Division

311
00:10:11.590 --> 00:10:13.210
were in Hispanic neighborhoods.

312
00:10:13.210 --> 00:10:15.230
But police say that's by chance.

313
00:10:15.230 --> 00:10:17.600
<v ->I do believe the data is objective.</v>

314
00:10:17.600 --> 00:10:20.220
We're simply using the crime numbers.

315
00:10:20.220 --> 00:10:21.300
The type of crime.

316
00:10:21.300 --> 00:10:22.176
The location.

317
00:10:22.176 --> 00:10:23.390 line:15% 
And

318
00:10:23.390 --> 00:10:24.380 line:15% 
the time

319
00:10:24.380 --> 00:10:25.470 line:15% 
of when we thought that crime,

320
00:10:25.470 --> 00:10:26.970 line:15% 
or when that crime was reported to us.

321
00:10:26.970 --> 00:10:30.730
It has nothing to do with anybody's ethnic makeup.

322
00:10:30.730 --> 00:10:32.300
Or the neighborhood.

323
00:10:32.300 --> 00:10:34.670
Or anything else.

324
00:10:34.670 --> 00:10:37.030
<v Issie>According to research by PredPol's co-founder,</v>

325
00:10:37.030 --> 00:10:39.430
their algorithms don't lead to more arrests of minorities

326
00:10:39.430 --> 00:10:40.920
than human policing.

327
00:10:40.920 --> 00:10:42.370
But that isn't exactly comforting

328
00:10:42.370 --> 00:10:44.340
to critics of traditional policing.

329
00:10:44.340 --> 00:10:47.580
<v ->The issue is that crime tends to happen in all areas.</v>

330
00:10:47.580 --> 00:10:48.950
But the problem is that

331
00:10:48.950 --> 00:10:50.860
marginalized groups are where police

332
00:10:50.860 --> 00:10:52.790
are usually congregating.

333
00:10:52.790 --> 00:10:55.610
<v Issie>And there is evidence of that disparity.</v>

334
00:10:55.610 --> 00:10:57.420
Today black and Hispanic Americans

335
00:10:57.420 --> 00:10:59.910
account for about a quarter of the total population.

336
00:10:59.910 --> 00:11:02.693
But make up more than half of the prison population.

337
00:11:03.640 --> 00:11:05.530
And there may not be a more infamous symbol

338
00:11:05.530 --> 00:11:06.550
of unjust policing

339
00:11:06.550 --> 00:11:07.800
than the Rodney King beating,

340
00:11:07.800 --> 00:11:08.770
which happened right here

341
00:11:08.770 --> 00:11:11.790
in LAPD's Foothill Division 27 years ago.

342
00:11:11.790 --> 00:11:13.430
That beating and the riots sparked

343
00:11:13.430 --> 00:11:15.200
by the acquittal of the officers involved,

344
00:11:15.200 --> 00:11:16.840
ignited a national conversation

345
00:11:16.840 --> 00:11:19.253
around police brutality that continues today.

346
00:11:20.580 --> 00:11:22.920
Members of the LAPD say that predictive policing

347
00:11:22.920 --> 00:11:25.690
is helping them move forward as an organization.

348
00:11:25.690 --> 00:11:27.640
<v ->Instead of being an occupying force,</v>

349
00:11:27.640 --> 00:11:29.790
like we were referred to once upon a time.

350
00:11:29.790 --> 00:11:31.610
Where there was a problem in the area,

351
00:11:31.610 --> 00:11:33.240
and we had to come in with tons of cops.

352
00:11:33.240 --> 00:11:34.790
We're tryna be more strategic.

353
00:11:34.790 --> 00:11:37.410
<v Issie>But activists of the Youth Justice Coalition</v>

354
00:11:37.410 --> 00:11:40.240
say that new technologies will only make things worse.

355
00:11:40.240 --> 00:11:44.180
<v ->It's basically allowing them to racially profile</v>

356
00:11:44.180 --> 00:11:45.280
black and brown youth.

357
00:11:45.280 --> 00:11:47.470
<v ->I live in a low income neighborhood.</v>

358
00:11:47.470 --> 00:11:49.980
Mostly black and brown people that live there.

359
00:11:49.980 --> 00:11:51.590
There's always police on my block.

360
00:11:51.590 --> 00:11:52.620
What ProdPol I think is doing

361
00:11:52.620 --> 00:11:56.180
is targeting communities that they already been targeting.

362
00:11:56.180 --> 00:11:57.840
<v Issie>Stop LAPD Spying Coalition</v>

363
00:11:57.840 --> 00:12:00.660
is still pursuing their lawsuit against the LAPD.

364
00:12:00.660 --> 00:12:03.650
And for now the most that concerned citizens can hope for

365
00:12:03.650 --> 00:12:05.880
is more transparency.

366
00:12:05.880 --> 00:12:06.840
<v ->If there was a dialog</v>

367
00:12:06.840 --> 00:12:08.080
between these police departments

368
00:12:08.080 --> 00:12:09.700
and the communities that they're serving,

369
00:12:09.700 --> 00:12:12.050
explaining to them how they're using these data points.

370
00:12:12.050 --> 00:12:13.810
How they're using this technology.

371
00:12:13.810 --> 00:12:15.000
It may go a long way.

372
00:12:15.000 --> 00:12:17.050
<v ->Predictive policing is here to stay.</v>

373
00:12:17.050 --> 00:12:18.660
And so what we should do is

374
00:12:18.660 --> 00:12:22.100
figure out how to identify the risks,

375
00:12:22.100 --> 00:12:24.690
see them coming and address them beforehand

376
00:12:24.690 --> 00:12:25.740
before it's too late.