WEBVTT

1
00:00:02.100 --> 00:00:05.210
We hear a lot about the bias AI can exacerbate.

2
00:00:05.210 --> 00:00:08.510
But AI can help organizations reduce bias too.

3
00:00:08.510 --> 00:00:11.190
Find out how when we talk with JoAnn Stonier,

4
00:00:11.190 --> 00:00:13.113
chief data officer at MasterCard.

5
00:00:14.190 --> 00:00:16.220
Welcome to "Me, Myself, and AI,"

6
00:00:16.220 --> 00:00:19.410
a podcast on artificial intelligence in business.

7
00:00:19.410 --> 00:00:20.930
Each episode, we introduce you

8
00:00:20.930 --> 00:00:23.280
to someone innovating with AI.

9
00:00:23.280 --> 00:00:26.140
I'm Sam Ransbotham, professor of information systems

10
00:00:26.140 --> 00:00:27.810
at Boston College.

11
00:00:27.810 --> 00:00:29.240
I'm also the guest editor

12
00:00:29.240 --> 00:00:32.100
for the "AI and Business Strategy" Big Idea program

13
00:00:32.100 --> 00:00:34.900
at MIT Sloan Management Review.

14
00:00:34.900 --> 00:00:38.800
And I'm Shervin Khodabandeh, senior partner with BCG.

15
00:00:38.800 --> 00:00:42.590
And I call it BCG's AI practice in North America.

16
00:00:42.590 --> 00:00:43.850
And together,

17
00:00:43.850 --> 00:00:48.090
MITSMR and BCG have been researching AI for five years,

18
00:00:48.090 --> 00:00:50.270
interviewing hundreds of practitioners

19
00:00:50.270 --> 00:00:52.300
and surveying thousands of companies

20
00:00:52.300 --> 00:00:55.080
on what it takes to build and to deploy

21
00:00:55.080 --> 00:00:58.280
and scale AI capabilities across the organization,

22
00:00:58.280 --> 00:01:00.843
and really transform the way organizations operate.

23
00:01:03.240 --> 00:01:05.370
Today we're talking with JoAnn Stonier.

24
00:01:05.370 --> 00:01:07.600
JoAnn is chief data officer at MasterCard.

25
00:01:07.600 --> 00:01:09.750
JoAnne, thanks for taking the time to talk with us today.

26
00:01:09.750 --> 00:01:10.710
Welcome.

27
00:01:10.710 --> 00:01:12.800
Thank you, happy to be here.

28
00:01:12.800 --> 00:01:15.480
JoAnn, let's start with your current role at MasterCard.

29
00:01:15.480 --> 00:01:18.030
Can you give us a quick overview of what you do?

30
00:01:18.030 --> 00:01:19.820
Sure, currently as you said,

31
00:01:19.820 --> 00:01:22.010
the chief data officer for the firm.

32
00:01:22.010 --> 00:01:24.240
And me and my team are responsible

33
00:01:24.240 --> 00:01:26.930
for ensuring that MasterCard's information assets

34
00:01:26.930 --> 00:01:28.810
are available for information

35
00:01:28.810 --> 00:01:31.920
while navigating current and future data risk.

36
00:01:31.920 --> 00:01:34.620
So my team has a very broad mandate.

37
00:01:34.620 --> 00:01:38.340
We work on helping the firm develop our data strategy.

38
00:01:38.340 --> 00:01:40.680
And then work on all the different aspects

39
00:01:40.680 --> 00:01:44.610
of data management including data governance, data quality,

40
00:01:44.610 --> 00:01:47.760
as well as enabling things like artificial intelligence,

41
00:01:47.760 --> 00:01:49.070
machine learning,

42
00:01:49.070 --> 00:01:51.820
and then we also help design and operate

43
00:01:51.820 --> 00:01:53.880
some of our enterprise data platforms.

44
00:01:53.880 --> 00:01:55.170
It's a very broad-based role.

45
00:01:55.170 --> 00:01:56.950
We work also on data compliance.

46
00:01:56.950 --> 00:01:58.750
And how do you embed compliance

47
00:01:58.750 --> 00:02:02.450
and responsible data practices right into product design.

48
00:02:02.450 --> 00:02:04.900
We start at the very beginning of data sourcing

49
00:02:04.900 --> 00:02:07.970
all the way through to product creation and enablement.

50
00:02:07.970 --> 00:02:09.100
It's a lot of fun.

51
00:02:09.100 --> 00:02:11.270
Because our show is about artificial intelligence

52
00:02:11.270 --> 00:02:13.040
let me pick up on that aspect.

53
00:02:13.040 --> 00:02:15.610
Can you give us examples of something you're excited about

54
00:02:15.610 --> 00:02:18.060
that MasterCard is doing with artificial intelligence?

55
00:02:18.060 --> 00:02:21.090
Oh my gosh, I've had so many conversations just this week

56
00:02:21.090 --> 00:02:23.060
about artificial intelligence.

57
00:02:23.060 --> 00:02:26.030
Most of them centered on minimization of bias

58
00:02:26.030 --> 00:02:28.660
as well as how do we build an inclusive future?

59
00:02:28.660 --> 00:02:30.630
The conversations that really excite me

60
00:02:30.630 --> 00:02:31.530
is how the whole firm

61
00:02:31.530 --> 00:02:33.970
is really getting behind this idea and notion.

62
00:02:33.970 --> 00:02:37.270
I've had conversations with our product development team

63
00:02:37.270 --> 00:02:39.520
and how do we develop a broad based playbook.

64
00:02:39.520 --> 00:02:42.650
So that everybody in the organization really understands

65
00:02:42.650 --> 00:02:44.820
how do you begin to really think about design

66
00:02:44.820 --> 00:02:46.220
at its very inception

67
00:02:46.220 --> 00:02:48.990
so that you're really thinking about inclusive concepts?

68
00:02:48.990 --> 00:02:50.100
We've had conversations

69
00:02:50.100 --> 00:02:51.730
with our people and capabilities team

70
00:02:51.730 --> 00:02:54.470
or what's a more commonly known as human resources.

71
00:02:54.470 --> 00:02:56.680
Talking about the skillsets of the future,

72
00:02:56.680 --> 00:02:58.690
how are we going to not only have

73
00:02:58.690 --> 00:03:00.870
a more inclusive workforce at MasterCard.

74
00:03:00.870 --> 00:03:02.290
And what do we need to do

75
00:03:02.290 --> 00:03:04.540
to provide education opportunities

76
00:03:04.540 --> 00:03:06.950
both inside the firm but also outside the firm.

77
00:03:06.950 --> 00:03:08.430
So that we can create

78
00:03:08.430 --> 00:03:10.530
the right kind of profile of individuals

79
00:03:10.530 --> 00:03:12.950
so that they have the skill sets that we need.

80
00:03:12.950 --> 00:03:14.610
But also how do we upscale,

81
00:03:14.610 --> 00:03:17.450
how do we really begin to create the opportunities

82
00:03:17.450 --> 00:03:19.660
to have the right kinds of conversations?

83
00:03:19.660 --> 00:03:23.400
We're also working really hard on our ethical AI process.

84
00:03:23.400 --> 00:03:25.410
So there's so many different aspects

85
00:03:25.410 --> 00:03:28.040
of what we're doing around artificial intelligence.

86
00:03:28.040 --> 00:03:32.240
Not just in our products and solutions in fraud and cyber,

87
00:03:32.240 --> 00:03:35.600
in general analytics and in biometric solutions

88
00:03:35.600 --> 00:03:37.093
around digital identity.

89
00:03:38.423 --> 00:03:41.870
It's really an interesting time to do this work.

90
00:03:41.870 --> 00:03:44.580
And to really do it in a way that I think needs to last

91
00:03:44.580 --> 00:03:45.973
for the generations ahead.

92
00:03:46.840 --> 00:03:48.240
JoAnn that's really interesting.

93
00:03:48.240 --> 00:03:50.720
I'm particularly excited

94
00:03:50.720 --> 00:03:54.080
that one of the things you talked about is use of AI

95
00:03:54.080 --> 00:03:56.230
to prevent bias.

96
00:03:56.230 --> 00:03:57.780
There's been a lot of conversations

97
00:03:57.780 --> 00:04:02.780
around the unintended bias of AI and how to manage it.

98
00:04:02.970 --> 00:04:05.250
But I heard you also referred to it

99
00:04:05.250 --> 00:04:08.930
as a tool that can actually help uncover biases.

100
00:04:08.930 --> 00:04:10.530
Can you comment more about that?

101
00:04:11.510 --> 00:04:13.110
Yeah, but it's hard, right?

102
00:04:13.110 --> 00:04:14.430
This is a lot of hard work.

103
00:04:14.430 --> 00:04:18.440
I do a lot of conversations both with academic institutions,

104
00:04:18.440 --> 00:04:20.940
other civil society organizations.

105
00:04:20.940 --> 00:04:23.850
Right, we're early days of AI and machine learning.

106
00:04:23.850 --> 00:04:26.939
I think we're probably at generation maybe 1.5

107
00:04:26.939 --> 00:04:28.920
heading into generation 2.

108
00:04:28.920 --> 00:04:31.280
But I think the events of this past year

109
00:04:31.280 --> 00:04:33.810
have taught us that we really need to pay attention

110
00:04:33.810 --> 00:04:38.430
to how we are designing products and solutions for society.

111
00:04:38.430 --> 00:04:41.980
And that our data sets are really important

112
00:04:41.980 --> 00:04:44.410
in what are we feeding into the machines

113
00:04:44.410 --> 00:04:47.280
and how do we design our algorithmic processes

114
00:04:47.280 --> 00:04:50.650
that we also were feeding into the intelligence

115
00:04:50.650 --> 00:04:52.350
and what is it going to learn from us?

116
00:04:52.350 --> 00:04:55.130
And so I try to explain to people

117
00:04:55.130 --> 00:04:57.490
our first generation of data analytics.

118
00:04:57.490 --> 00:05:01.920
We were creating algorithms or analytic questions

119
00:05:01.920 --> 00:05:05.990
that really asked the question of, this is the question,

120
00:05:05.990 --> 00:05:09.330
if this condition exists in the dataset

121
00:05:09.330 --> 00:05:11.270
then do something, right?

122
00:05:11.270 --> 00:05:13.730
So we were looking for the condition to exist

123
00:05:13.730 --> 00:05:16.069
in the data set and then we were acting-

124
00:05:16.069 --> 00:05:17.420
Rules-based.

125
00:05:17.420 --> 00:05:19.420
Right, rules-based, exactly.

126
00:05:19.420 --> 00:05:21.180
Artificial intelligence and machine learning

127
00:05:21.180 --> 00:05:23.820
actually flips that a little bit on the head.

128
00:05:23.820 --> 00:05:26.100
Instead we take the dataset

129
00:05:26.100 --> 00:05:28.450
and say: what does the data tell us?

130
00:05:28.450 --> 00:05:29.960
What does the data tell us

131
00:05:29.960 --> 00:05:32.210
and what does the machine then learn from that?

132
00:05:32.210 --> 00:05:33.610
Well, the challenge of that though

133
00:05:33.610 --> 00:05:36.260
is if you don't understand what's in the data, right?

134
00:05:36.260 --> 00:05:38.220
And the condition of that data

135
00:05:38.220 --> 00:05:40.180
as you bring it into the inquiry,

136
00:05:40.180 --> 00:05:43.430
and you don't understand how those two things fit,

137
00:05:43.430 --> 00:05:46.510
then you wind up with a biased perspective.

138
00:05:46.510 --> 00:05:50.070
Now, it's unintended, it's inherent, right, to your point.

139
00:05:50.070 --> 00:05:53.180
But nevertheless, it's something that we have to back up

140
00:05:53.180 --> 00:05:57.860
and readjust our lenses as we begin to look at our processes

141
00:05:57.860 --> 00:06:00.120
around artificial intelligence.

142
00:06:00.120 --> 00:06:02.000
So we start with the data sets.

143
00:06:02.000 --> 00:06:04.880
And we understand that data sets are going to have

144
00:06:04.880 --> 00:06:06.780
all sorts of bias in them and that's okay.

145
00:06:06.780 --> 00:06:08.690
First of all, bias always gets

146
00:06:09.601 --> 00:06:12.440
kind of a very gripping reaction, right?

147
00:06:12.440 --> 00:06:13.770
I use the example all the time,

148
00:06:13.770 --> 00:06:17.120
if you go back to like the 1910 voter roles

149
00:06:17.120 --> 00:06:18.750
in the United States,

150
00:06:18.750 --> 00:06:20.480
that's a valid data set.

151
00:06:20.480 --> 00:06:23.530
You may use that for whatever purpose you may have

152
00:06:23.530 --> 00:06:28.350
for evaluating something that happened in 1910 or 1911.

153
00:06:28.350 --> 00:06:31.400
But you need to know that inherently,

154
00:06:31.400 --> 00:06:34.770
that data set is going to miss women.

155
00:06:34.770 --> 00:06:37.710
It's going to be missing people of color.

156
00:06:37.710 --> 00:06:40.630
It's going to be missing parts of society.

157
00:06:40.630 --> 00:06:43.020
As long as you know that

158
00:06:43.020 --> 00:06:47.440
then you can design an inquiry that is fit for purpose.

159
00:06:47.440 --> 00:06:50.520
The problem is if you don't remember that

160
00:06:50.520 --> 00:06:52.670
or you're not mindful of that,

161
00:06:52.670 --> 00:06:57.160
then you have an inquiry that's gonna learn off of a dataset

162
00:06:57.160 --> 00:07:01.010
that is missing characteristics that's going to be important

163
00:07:01.010 --> 00:07:03.030
to whatever that other inquiry is.

164
00:07:03.030 --> 00:07:04.890
Those are some of the ways that I think

165
00:07:04.890 --> 00:07:07.500
we can actually begin to design a better future.

166
00:07:07.500 --> 00:07:09.980
But it means really being very mindful

167
00:07:09.980 --> 00:07:13.210
of what's inherent in the dataset,

168
00:07:13.210 --> 00:07:17.220
what's there, what's missing, what also can be imputed.

169
00:07:17.220 --> 00:07:18.620
We can talk about that,

170
00:07:18.620 --> 00:07:21.570
what kind of variables can be created by the machine, right?

171
00:07:21.570 --> 00:07:23.650
It can be imputed as well.

172
00:07:23.650 --> 00:07:25.130
And so all of those things

173
00:07:25.130 --> 00:07:26.900
are things that we're looking at at MasterCard.

174
00:07:26.900 --> 00:07:29.760
So we have a very specific framework that we're using

175
00:07:29.760 --> 00:07:31.660
around ethical AI.

176
00:07:31.660 --> 00:07:34.190
But then we're really getting into the nitty gritty

177
00:07:34.190 --> 00:07:36.580
around different kinds of contexts.

178
00:07:36.580 --> 00:07:37.730
So I wanna build on that

179
00:07:37.730 --> 00:07:40.280
because I think it's a very critical topic

180
00:07:40.280 --> 00:07:42.130
to so many of our listeners.

181
00:07:42.130 --> 00:07:45.080
So you talked about as long as you know

182
00:07:45.080 --> 00:07:46.800
you could look for it.

183
00:07:46.800 --> 00:07:48.740
But then the things you don't know

184
00:07:48.740 --> 00:07:51.470
and then you'd find out in the hindsight,

185
00:07:51.470 --> 00:07:53.640
how do you prepare for those?

186
00:07:53.640 --> 00:07:55.810
Well, I think you have to be aware

187
00:07:55.810 --> 00:07:58.750
again, for what purpose to what end.

188
00:07:58.750 --> 00:08:01.160
And now let's go look at the datasets.

189
00:08:01.160 --> 00:08:05.410
And this is where you've gotta backup the data food chain

190
00:08:05.410 --> 00:08:07.640
to understand your data lineage.

191
00:08:07.640 --> 00:08:10.680
Understand the quality, the consistency

192
00:08:10.680 --> 00:08:15.460
of the data sets that are going into the analysis.

193
00:08:15.460 --> 00:08:16.610
And then this is not easy.

194
00:08:16.610 --> 00:08:19.030
So nothing I'm saying, I want everybody to hear me clearly

195
00:08:19.030 --> 00:08:20.510
nothing I'm saying is easy, right?

196
00:08:20.510 --> 00:08:23.170
Everything requires a lot of scrutiny and time.

197
00:08:23.170 --> 00:08:25.810
But that's easiest if you're doing it yourself.

198
00:08:25.810 --> 00:08:26.950
But if a vendor is doing this

199
00:08:26.950 --> 00:08:29.560
and is presenting this to you as a combined solution,

200
00:08:29.560 --> 00:08:31.150
you need to be even more curious

201
00:08:31.150 --> 00:08:34.170
about what is going into this recipe.

202
00:08:34.170 --> 00:08:38.040
Every one of these elements becomes super important, why?

203
00:08:38.040 --> 00:08:39.980
Because if we do it when we're building it,

204
00:08:39.980 --> 00:08:42.510
those of us who are the data scientists and the data expert,

205
00:08:42.510 --> 00:08:44.700
the data designers, right?

206
00:08:44.700 --> 00:08:47.630
We understand it because when we're gonna be looking at

207
00:08:47.630 --> 00:08:50.300
the information models and the outputs as they come out

208
00:08:50.300 --> 00:08:51.370
in the first generations,

209
00:08:51.370 --> 00:08:52.850
we're gonna be looking for the model drift.

210
00:08:52.850 --> 00:08:55.910
We're gonna be looking to see: is this an accurate output?

211
00:08:55.910 --> 00:08:58.380
Is this truly the result

212
00:08:58.380 --> 00:09:01.900
or is it because of some inequity in the input?

213
00:09:01.900 --> 00:09:04.550
Which is okay if it's accurate, that's okay.

214
00:09:04.550 --> 00:09:06.360
Because the machine's gonna then start learning

215
00:09:06.360 --> 00:09:08.930
and we're gonna put more data and more information.

216
00:09:08.930 --> 00:09:11.630
But if it's incorrect and we haven't caught it

217
00:09:11.630 --> 00:09:15.160
it could get amplified to the incorrect result.

218
00:09:15.160 --> 00:09:16.390
And it could have,

219
00:09:16.390 --> 00:09:18.180
and this is what we also really care about

220
00:09:18.180 --> 00:09:20.280
<v ->that's when we get to the output analysis -</v>

221
00:09:20.280 --> 00:09:22.810
it could amplify really bad outcomes.

222
00:09:22.810 --> 00:09:26.360
And the outcomes can be significant or insignificant

223
00:09:26.360 --> 00:09:29.460
depending upon if it's being designed for an individual

224
00:09:29.460 --> 00:09:32.100
or if it's being designed for groups.

225
00:09:32.100 --> 00:09:33.430
Or if it's just being something...

226
00:09:33.430 --> 00:09:35.230
I talk about how I suspect

227
00:09:35.230 --> 00:09:37.550
that when the airlines were first modeling

228
00:09:37.550 --> 00:09:39.740
the size of the overhead bins,

229
00:09:39.740 --> 00:09:42.630
it was on very average heights, okay?

230
00:09:42.630 --> 00:09:45.350
But the average height did not include enough women.

231
00:09:45.350 --> 00:09:48.740
How many ladies climb on seats to get their luggage

232
00:09:48.740 --> 00:09:50.320
at the very back of the bend

233
00:09:50.320 --> 00:09:52.410
or need to rely on one of you nice gentlemen

234
00:09:52.410 --> 00:09:56.160
to grab our overhead suitcases down to assist us, right?

235
00:09:56.160 --> 00:09:59.350
Averages sometimes are the worst result

236
00:09:59.350 --> 00:10:01.660
because they're not fit for purpose.

237
00:10:01.660 --> 00:10:05.370
I think you're talking about injecting an entire new layer

238
00:10:05.370 --> 00:10:08.318
in the data analysis or data usage process, right?

239
00:10:08.318 --> 00:10:09.500
Maybe. (chuckles)

240
00:10:09.500 --> 00:10:12.340
Because, was it not the case, Sam

241
00:10:12.340 --> 00:10:16.980
that 20 years ago people thought of data as neutrals.

242
00:10:16.980 --> 00:10:19.970
Like data is neutral, you use what you want.

243
00:10:19.970 --> 00:10:21.820
But data could be potentially harmful.

244
00:10:21.820 --> 00:10:23.830
And I think what you're talking about is,

245
00:10:23.830 --> 00:10:27.160
you're trying to create a mindset shift

246
00:10:27.160 --> 00:10:31.040
that says it's not just data comes in, answers go out.

247
00:10:31.040 --> 00:10:33.330
But you have to be mindful of why you're using it,

248
00:10:33.330 --> 00:10:35.240
how you're using it.

249
00:10:35.240 --> 00:10:36.700
And this mindset shift

250
00:10:36.700 --> 00:10:38.680
just permeates not just your organization

251
00:10:38.680 --> 00:10:40.420
but also you're saying your vendors

252
00:10:40.420 --> 00:10:42.680
and other parts of the ecosystem.

253
00:10:42.680 --> 00:10:44.360
It's a huge undertaking.

254
00:10:44.360 --> 00:10:46.110
It is a huge undertaking.

255
00:10:46.110 --> 00:10:47.930
So those are some challenging aspects,

256
00:10:47.930 --> 00:10:48.830
what's rewarding?

257
00:10:48.830 --> 00:10:50.210
What kind of exciting things happen?

258
00:10:50.210 --> 00:10:51.797
These seem a little bit sort of like,

259
00:10:51.797 --> 00:10:53.380
"Oh gosh, we have to worry about this,

260
00:10:53.380 --> 00:10:55.400
have to worry about that!"

261
00:10:55.400 --> 00:10:56.840
Get us excited about the fun part.

262
00:10:56.840 --> 00:10:59.520
Oh, I think all of this is fun, I do.

263
00:10:59.520 --> 00:11:00.560
But I'm a data geek, right?

264
00:11:00.560 --> 00:11:04.160
So what's fun is actually seeing my firm

265
00:11:04.160 --> 00:11:05.680
come to life around this.

266
00:11:05.680 --> 00:11:08.470
These are actually very exciting conversations

267
00:11:08.470 --> 00:11:09.540
inside of the firm.

268
00:11:09.540 --> 00:11:10.640
I don't mean to make them sound

269
00:11:10.640 --> 00:11:13.140
like just risk conversations at all.

270
00:11:13.140 --> 00:11:15.320
It's intellectually challenging work.

271
00:11:15.320 --> 00:11:17.090
But we all agree that we think

272
00:11:17.090 --> 00:11:19.380
it's going to bring us to better places.

273
00:11:19.380 --> 00:11:21.580
And already is bringing us to better places

274
00:11:21.580 --> 00:11:22.493
in product design.

275
00:11:23.370 --> 00:11:26.230
You talked about the workforce aspect of it

276
00:11:26.230 --> 00:11:29.380
and that technology is necessary

277
00:11:29.380 --> 00:11:31.720
but is as far from sufficient.

278
00:11:31.720 --> 00:11:33.930
We heard from one of our other speakers,

279
00:11:33.930 --> 00:11:37.400
the president of 1-800-Flowers about learning quotient,

280
00:11:37.400 --> 00:11:39.700
the desire to wanna learn.

281
00:11:39.700 --> 00:11:41.640
How do you bring all of that together

282
00:11:41.640 --> 00:11:44.100
in terms of the future of the workforce

283
00:11:44.100 --> 00:11:45.870
that you guys are looking at

284
00:11:45.870 --> 00:11:49.290
and the kind of talent and skillset and attributes

285
00:11:49.290 --> 00:11:52.210
that are going to be successful in different roles?

286
00:11:52.210 --> 00:11:55.720
Are you taking actions in that direction?

287
00:11:55.720 --> 00:11:56.700
We really look for people

288
00:11:56.700 --> 00:11:59.710
who also can design controls and processes

289
00:11:59.710 --> 00:12:03.570
and make sure that we can navigate those.

290
00:12:03.570 --> 00:12:06.470
We live in a world of connected ecosystems

291
00:12:06.470 --> 00:12:08.730
and so we're only as good as our partners.

292
00:12:08.730 --> 00:12:10.860
But it means those handshakes, right?

293
00:12:10.860 --> 00:12:13.800
Those connections are super important to understand.

294
00:12:13.800 --> 00:12:16.490
And so how do you make those inquiries

295
00:12:16.490 --> 00:12:18.040
of other organizations

296
00:12:18.040 --> 00:12:20.850
so that you can create those connected ecosystems

297
00:12:20.850 --> 00:12:23.440
that are only going to grow in size and scale

298
00:12:23.440 --> 00:12:24.630
for the future?

299
00:12:24.630 --> 00:12:25.990
And how do we help design those?

300
00:12:25.990 --> 00:12:28.250
How do we be the leaders in designing some of those

301
00:12:28.250 --> 00:12:30.190
is really, really important for us.

302
00:12:30.190 --> 00:12:32.760
So it really is this meshing.

303
00:12:32.760 --> 00:12:35.260
I talked about being like a three-sided triangle

304
00:12:35.260 --> 00:12:36.780
of data skills-

305
00:12:36.780 --> 00:12:40.070
Is there a different kind?
(all laughing)

306
00:12:40.070 --> 00:12:41.270
But we sit in the middle, right?

307
00:12:41.270 --> 00:12:42.610
We sit in the middle of the business.

308
00:12:42.610 --> 00:12:45.020
We sit with technology skills.

309
00:12:45.020 --> 00:12:47.160
As well as there's an awful lot of policy

310
00:12:47.160 --> 00:12:49.060
that we need to understand often.

311
00:12:49.060 --> 00:12:51.140
It's the other design constraint

312
00:12:51.140 --> 00:12:55.210
of where are law's evolving, where is the restrictions?

313
00:12:55.210 --> 00:12:58.410
The other type of restrictions that we need to know.

314
00:12:58.410 --> 00:13:00.040
Is it gonna be on soil?

315
00:13:00.040 --> 00:13:01.170
Can we not use the data

316
00:13:01.170 --> 00:13:03.390
because it's contractually restricted?

317
00:13:03.390 --> 00:13:06.500
Or it's restricted because it's a certain type of data

318
00:13:06.500 --> 00:13:08.010
it's financial data, right?

319
00:13:08.010 --> 00:13:11.000
It only can be used for one specific type of purpose.

320
00:13:11.000 --> 00:13:14.380
Or it can only be used at an aggregated level for example.

321
00:13:14.380 --> 00:13:16.080
Or it can be used at a segment level.

322
00:13:16.080 --> 00:13:19.010
Those types of restrictions as well at that compliance level

323
00:13:19.010 --> 00:13:20.250
also needs to be understood.

324
00:13:20.250 --> 00:13:21.770
So sitting in the middle of the middle

325
00:13:21.770 --> 00:13:22.770
is sometimes really hard.

326
00:13:22.770 --> 00:13:24.300
So that's why you have to be able to admit

327
00:13:24.300 --> 00:13:26.753
you don't know one aspect of that triangle.

328
00:13:27.660 --> 00:13:29.740
So you mentioned the word design,

329
00:13:29.740 --> 00:13:32.810
a couple of times at least.
(JoAnne laughs)

330
00:13:32.810 --> 00:13:34.560
Here it comes.
(Sam laughs)

331
00:13:34.560 --> 00:13:37.147
Can you tell us little about how you got to your role?

332
00:13:37.147 --> 00:13:39.060
Because I think the design thinking

333
00:13:39.060 --> 00:13:41.110
is coming from your background.

334
00:13:41.110 --> 00:13:44.053
Can you share with us how you got to where you are?

335
00:13:44.890 --> 00:13:45.723
I think it's probably easier

336
00:13:45.723 --> 00:13:46.710
if I just go quickly backwards.

337
00:13:46.710 --> 00:13:49.020
So prior to being the chief data officer

338
00:13:49.020 --> 00:13:51.640
I was the chief privacy officer for the firm.

339
00:13:51.640 --> 00:13:53.660
Which was a lot of fun.

340
00:13:53.660 --> 00:13:55.330
I enjoyed that role immensely.

341
00:13:55.330 --> 00:13:58.700
I helped the company become GDPR compliant.

342
00:13:58.700 --> 00:14:02.110
Which really was, I think the moment in time

343
00:14:02.110 --> 00:14:05.100
when many many companies were coming to the place

344
00:14:05.100 --> 00:14:08.570
where they needed to operationalize a lot of data risk.

345
00:14:08.570 --> 00:14:10.060
And we had been on that journey.

346
00:14:10.060 --> 00:14:12.200
We had a whole process called privacy by design.

347
00:14:12.200 --> 00:14:13.890
We still use that process.

348
00:14:13.890 --> 00:14:15.933
Lots of companies use that phrase

349
00:14:15.933 --> 00:14:17.360
because it's a regulatory phrase.

350
00:14:17.360 --> 00:14:19.810
But we had already been

351
00:14:19.810 --> 00:14:21.540
looking at our products and solutions.

352
00:14:21.540 --> 00:14:24.410
To try to understand how could we embed privacy and security

353
00:14:24.410 --> 00:14:27.350
into their very design and into their fabric.

354
00:14:27.350 --> 00:14:29.680
And we still do that to this day.

355
00:14:29.680 --> 00:14:31.810
But as we were doing the compliance work

356
00:14:31.810 --> 00:14:35.820
and planning it out for the GDPR, the European privacy law,

357
00:14:35.820 --> 00:14:39.460
we were realizing we were gonna need additional platforms

358
00:14:39.460 --> 00:14:40.990
and systems to be built.

359
00:14:40.990 --> 00:14:42.080
And as I was doing that

360
00:14:42.080 --> 00:14:43.897
I kept saying to anyone who had listened,

361
00:14:43.897 --> 00:14:45.330
"Who's gonna own all the data

362
00:14:45.330 --> 00:14:47.370
because I really need somebody to speak with?"

363
00:14:47.370 --> 00:14:48.697
And they just kept patting me on the head saying,

364
00:14:48.697 --> 00:14:52.190
"You just keep designing everything and it'll be fine."

365
00:14:52.190 --> 00:14:55.110
So here we are that I'm the first chief data officer.

366
00:14:55.110 --> 00:14:58.160
But it was kind of an expansion

367
00:14:58.160 --> 00:14:59.920
and then a severing of a lot of the work

368
00:14:59.920 --> 00:15:01.170
that I had been doing.

369
00:15:01.170 --> 00:15:03.070
But I came to be the chief privacy officer.

370
00:15:03.070 --> 00:15:04.840
I was previously the chief privacy officer

371
00:15:04.840 --> 00:15:06.040
for American Express.

372
00:15:06.040 --> 00:15:09.930
I came to that role after a pretty good size career

373
00:15:09.930 --> 00:15:11.930
at American Express as well.

374
00:15:11.930 --> 00:15:13.390
Working in a variety of roles

375
00:15:13.390 --> 00:15:14.840
I understood how that firm works.

376
00:15:14.840 --> 00:15:16.040
So financial services

377
00:15:16.040 --> 00:15:18.130
is something I've been doing for a while.

378
00:15:18.130 --> 00:15:21.780
But the design piece comes in because of 9/11.

379
00:15:21.780 --> 00:15:24.430
I had the misfortune of being downtown that day.

380
00:15:24.430 --> 00:15:27.590
And the American Express building is right across the street

381
00:15:27.590 --> 00:15:30.223
from was the World Trade Center.

382
00:15:31.163 --> 00:15:34.010
So saw a lot and lost colleagues that day.

383
00:15:34.010 --> 00:15:36.630
And we were all relocated for several months

384
00:15:36.630 --> 00:15:38.470
while the building was repaired.

385
00:15:38.470 --> 00:15:41.047
And so on New Years of the next, I don't know

386
00:15:41.047 --> 00:15:44.740
it was probably like January 4th actually of 2002.

387
00:15:44.740 --> 00:15:46.350
I was thinking, well, there's a lot of people

388
00:15:46.350 --> 00:15:48.760
who were not alive that New Year's.

389
00:15:48.760 --> 00:15:50.870
And I thought, well, what can I do?

390
00:15:50.870 --> 00:15:51.990
What haven't I done in my life

391
00:15:51.990 --> 00:15:54.490
that I really want to go accomplish?

392
00:15:54.490 --> 00:15:56.990
And so I decided to go back to design school.

393
00:15:56.990 --> 00:15:58.010
This is after law school.

394
00:15:58.010 --> 00:16:00.580
So I got my law transcripts, were part of the process.

395
00:16:00.580 --> 00:16:03.370
Which was fun getting them to send my transcripts

396
00:16:03.370 --> 00:16:04.570
to a design school.

397
00:16:04.570 --> 00:16:06.090
But I got a design degree.

398
00:16:06.090 --> 00:16:07.120
And I'm really glad I did

399
00:16:07.120 --> 00:16:09.590
because it's made me a better business person.

400
00:16:09.590 --> 00:16:12.070
And it's given me a whole different way of thinking

401
00:16:12.070 --> 00:16:13.800
about problem solving.

402
00:16:13.800 --> 00:16:15.010
And then I've had the good fortune.

403
00:16:15.010 --> 00:16:17.700
So I do, yes I do interior design work.

404
00:16:17.700 --> 00:16:20.180
And yes I can help you with your kitchen and your bathroom

405
00:16:20.180 --> 00:16:22.330
and all the other things you wanna do during COVID.

406
00:16:22.330 --> 00:16:24.490
I've helped several colleagues pick out tile

407
00:16:24.490 --> 00:16:25.370
and other things.

408
00:16:25.370 --> 00:16:26.540
All right, we'll talk after the podcast about that.

409
00:16:26.540 --> 00:16:29.130
We can talk after the podcast about that.

410
00:16:29.130 --> 00:16:30.700
But I also then had the good fortune

411
00:16:30.700 --> 00:16:33.080
to meet the Dean at Pratt Institute.

412
00:16:33.080 --> 00:16:36.120
And I teach in their design management master's program.

413
00:16:36.120 --> 00:16:37.490
And I teach business strategy

414
00:16:37.490 --> 00:16:39.720
and I've taught other courses in that program.

415
00:16:39.720 --> 00:16:41.010
And that's also helped me

416
00:16:41.010 --> 00:16:43.190
really evolve my design thinking as well.

417
00:16:43.190 --> 00:16:45.130
And so all of these things,

418
00:16:45.130 --> 00:16:47.680
yes, I have a law degree, yes, I'm a data person.

419
00:16:47.680 --> 00:16:50.270
But having the design thinking as well

420
00:16:50.270 --> 00:16:52.460
really makes you not think of things as problems.

421
00:16:52.460 --> 00:16:54.210
So just constraints around

422
00:16:54.210 --> 00:16:56.010
which you have to design things

423
00:16:56.010 --> 00:16:57.410
and that they will shift over time.

424
00:16:57.410 --> 00:16:59.550
So it's no different than the electric is over here

425
00:16:59.550 --> 00:17:00.680
and the plumbing is over here

426
00:17:00.680 --> 00:17:02.490
and you only have so much space.

427
00:17:02.490 --> 00:17:04.480
So how can you utilize this, right?

428
00:17:04.480 --> 00:17:05.870
It's the same thing for,

429
00:17:05.870 --> 00:17:07.700
I can only use the data in this way

430
00:17:07.700 --> 00:17:11.380
and I want to achieve an outcome, how do I do that?

431
00:17:11.380 --> 00:17:14.750
And so it is the same kind of strategic toolkit.

432
00:17:14.750 --> 00:17:16.527
You just kind of flip it for,

433
00:17:16.527 --> 00:17:19.669
"Okay, well, the powder room has to be only this big."

434
00:17:19.669 --> 00:17:22.210
(chuckles) So that's one challenge.

435
00:17:22.210 --> 00:17:27.210
Or the outcome is to design a look alike model for fraud

436
00:17:27.440 --> 00:17:30.660
that only utilizes synthetic data, right?

437
00:17:30.660 --> 00:17:32.150
Or something along those lines.

438
00:17:32.150 --> 00:17:34.570
But it does kind of give you a little bit of a can-do spirit

439
00:17:34.570 --> 00:17:37.150
because you figure something has to be possible.

440
00:17:37.150 --> 00:17:38.730
Because there's raw material in the world.

441
00:17:38.730 --> 00:17:41.180
So that's kind of how I approach things.

442
00:17:41.180 --> 00:17:42.310
Very well said.

443
00:17:42.310 --> 00:17:44.230
That's a very weird triangle.

444
00:17:44.230 --> 00:17:45.976
Seems the triangle is a very weird-

445
00:17:45.976 --> 00:17:49.600
My triangle is probably very weird, I get a lot.

446
00:17:49.600 --> 00:17:52.270
Well, that's a bit of a theme though.

447
00:17:52.270 --> 00:17:53.510
Other people we've talked to

448
00:17:53.510 --> 00:17:56.290
they're bringing in these experiences from other places.

449
00:17:56.290 --> 00:17:59.430
And very often they're not super technical things.

450
00:17:59.430 --> 00:18:00.740
We're talking about artificial intelligence

451
00:18:00.740 --> 00:18:03.470
and you might think that it would quickly go down a path

452
00:18:03.470 --> 00:18:04.930
of extreme technical.

453
00:18:04.930 --> 00:18:09.890
But I think it's just as likely to go down a path of design.

454
00:18:09.890 --> 00:18:11.710
Yeah, I know and what we find on my team

455
00:18:11.710 --> 00:18:14.013
and it generally is that if you find people

456
00:18:14.013 --> 00:18:15.560
that have kind of that combination

457
00:18:15.560 --> 00:18:17.030
of right brain left brain.

458
00:18:17.030 --> 00:18:20.050
So I have lawyers and engineers on my team,

459
00:18:20.050 --> 00:18:21.540
they're both degrees.

460
00:18:21.540 --> 00:18:23.200
It's a really interesting match

461
00:18:23.200 --> 00:18:27.210
of being able to translate the business to the technical,

462
00:18:27.210 --> 00:18:28.870
the technical to the business, right?

463
00:18:28.870 --> 00:18:31.590
Translating the regulatory to the business

464
00:18:31.590 --> 00:18:32.920
or the regulatory to the legal.

465
00:18:32.920 --> 00:18:36.010
It kind of works as a giant translator role.

466
00:18:36.010 --> 00:18:37.960
And if you think about it, that's kind of the moment

467
00:18:37.960 --> 00:18:39.810
that we're in right now.

468
00:18:39.810 --> 00:18:44.420
Is the ability to be fluid in translating concepts

469
00:18:44.420 --> 00:18:46.920
from one domain into another.

470
00:18:46.920 --> 00:18:48.140
I think it works.

471
00:18:48.140 --> 00:18:50.620
But I do think that some of those competencies

472
00:18:50.620 --> 00:18:52.640
are gonna be equally important or more important

473
00:18:52.640 --> 00:18:56.560
as we continue to evolve, as we begin to develop.

474
00:18:56.560 --> 00:18:59.650
They'll be the core skill sets that we've always had,

475
00:18:59.650 --> 00:19:02.150
leadership skills and problem-solving skills

476
00:19:02.150 --> 00:19:03.250
and analytics skills.

477
00:19:03.250 --> 00:19:05.620
But I do think that the ability to translate that

478
00:19:05.620 --> 00:19:08.420
and then also be able to derive from the output, right?

479
00:19:08.420 --> 00:19:11.090
Whether it's from a dashboard or reports or whatever

480
00:19:11.090 --> 00:19:12.140
I think that's also gonna be

481
00:19:12.140 --> 00:19:13.340
really important business skills

482
00:19:13.340 --> 00:19:15.770
for any businessman or woman.

483
00:19:15.770 --> 00:19:16.800
You said right now,

484
00:19:16.800 --> 00:19:18.820
you said like that's the place we are right now.

485
00:19:18.820 --> 00:19:20.040
What do you think has caused that?

486
00:19:20.040 --> 00:19:22.140
What's making that so important right now?

487
00:19:23.190 --> 00:19:25.120
I don't know that I was emphasizing that.

488
00:19:25.120 --> 00:19:27.120
But I think when I look over the past year

489
00:19:27.120 --> 00:19:28.090
and I think about,

490
00:19:28.090 --> 00:19:31.870
we're just on the year mark of COVID for us.

491
00:19:31.870 --> 00:19:33.800
I can remember when we went virtual.

492
00:19:33.800 --> 00:19:35.950
I remember so many of the customer and client calls

493
00:19:35.950 --> 00:19:38.460
that I was helping to fill at the time.

494
00:19:38.460 --> 00:19:41.310
Companies were in a moment, right?

495
00:19:41.310 --> 00:19:43.400
So many merchants were going online.

496
00:19:43.400 --> 00:19:44.810
If they didn't have a huge presence

497
00:19:44.810 --> 00:19:46.410
they needed to create one.

498
00:19:46.410 --> 00:19:48.630
Business models were really shaken.

499
00:19:48.630 --> 00:19:50.200
But everybody was looking for data

500
00:19:50.200 --> 00:19:54.170
to try to help make some decisions, right?

501
00:19:54.170 --> 00:19:55.560
And they needed guidance,

502
00:19:55.560 --> 00:19:57.700
if they didn't have the data, they were looking for data.

503
00:19:57.700 --> 00:19:58.680
If they had the data,

504
00:19:58.680 --> 00:20:01.000
they were looking for guidance on how to interpret it.

505
00:20:01.000 --> 00:20:03.050
If they knew how to interpret it,

506
00:20:03.050 --> 00:20:05.300
they were looking for lookalike information

507
00:20:05.300 --> 00:20:07.890
to make sure they were making the right decisions.

508
00:20:07.890 --> 00:20:10.410
So if you just look at that maturity curve

509
00:20:10.410 --> 00:20:12.730
we have been thrust forward in time.

510
00:20:12.730 --> 00:20:14.150
So I think it would have been happening

511
00:20:14.150 --> 00:20:17.160
maybe in a less compressed way.

512
00:20:17.160 --> 00:20:19.050
I think we've been in a time and compression

513
00:20:19.050 --> 00:20:21.700
where data and digital has mattered.

514
00:20:21.700 --> 00:20:23.820
And so we can't undo that.

515
00:20:23.820 --> 00:20:27.410
And so I think these skillsets matter more and more

516
00:20:27.410 --> 00:20:29.810
just because of the compression we've been under

517
00:20:29.810 --> 00:20:30.820
as a society.

518
00:20:30.820 --> 00:20:33.510
And so we're going to see that

519
00:20:33.510 --> 00:20:37.430
be part of what this next normal or next generation

520
00:20:37.430 --> 00:20:39.040
because I don't really like the word normal.

521
00:20:39.040 --> 00:20:41.640
Next generation, this is now part and parcel

522
00:20:41.640 --> 00:20:43.420
of how we interact with each other.

523
00:20:43.420 --> 00:20:47.290
We will never go back to just being in person.

524
00:20:47.290 --> 00:20:49.740
This kind of connecting digitally

525
00:20:49.740 --> 00:20:53.480
will always now be part of our mandate, right?

526
00:20:53.480 --> 00:20:54.700
And part of the toolbox.

527
00:20:54.700 --> 00:20:57.820
So I just think that there will be more of a need

528
00:20:57.820 --> 00:21:00.310
to be able to interpret and to utilize

529
00:21:00.310 --> 00:21:03.050
than we had a year ago.

530
00:21:03.050 --> 00:21:05.700
That seems like your superpower is taking some situation

531
00:21:05.700 --> 00:21:08.850
whether it's 911 and turning it into a design school.

532
00:21:08.850 --> 00:21:10.900
Or taking a pandemic and turning it

533
00:21:10.900 --> 00:21:13.340
into a more fluid approach to things.

534
00:21:13.340 --> 00:21:15.500
It sounds like your super power in many cases.

535
00:21:15.500 --> 00:21:17.080
Thank you. (chuckles)

536
00:21:17.080 --> 00:21:18.293
I'll take that one.

537
00:21:20.280 --> 00:21:21.860
JoAnne, thank you so much for talking to us.

538
00:21:21.860 --> 00:21:23.600
You've really picked up on a lot of points.

539
00:21:23.600 --> 00:21:24.810
We've covered a lot of ground.

540
00:21:24.810 --> 00:21:26.660
Thanks for taking the time to talk with us today.

541
00:21:26.660 --> 00:21:27.910
Yeah, thank you so much.

542
00:21:27.910 --> 00:21:29.873
You're welcome. Thanks for having me.

543
00:21:29.873 --> 00:21:32.456
(gentle music)

544
00:21:35.780 --> 00:21:36.910
Trevor, that was great.

545
00:21:36.910 --> 00:21:38.610
JoAnne covered a lot today.

546
00:21:38.610 --> 00:21:41.060
What struck you as particularly interesting?

547
00:21:41.060 --> 00:21:44.730
A lot of things struck me in a positive way.

548
00:21:44.730 --> 00:21:47.810
I liked how she talked about design thinking.

549
00:21:47.810 --> 00:21:51.180
She talked about emotional quotient.

550
00:21:51.180 --> 00:21:54.410
And I actually think that's been part of her secret sauce

551
00:21:54.410 --> 00:21:55.910
and part of MasterCard success

552
00:21:55.910 --> 00:21:58.430
with these kinds of initiatives.

553
00:21:58.430 --> 00:21:59.680
Like if you look in the industry,

554
00:21:59.680 --> 00:22:03.890
the CDO role is actually somewhat precarious role.

555
00:22:03.890 --> 00:22:06.990
Because there's very successful CDOs

556
00:22:06.990 --> 00:22:11.080
and there's many CDOs that just don't have that vision

557
00:22:11.080 --> 00:22:13.140
or that ability or that autonomy

558
00:22:13.140 --> 00:22:16.570
or sort of that multi-sided triangle as she talked about.

559
00:22:16.570 --> 00:22:18.440
And so they don't succeed

560
00:22:18.440 --> 00:22:20.035
and they last for a couple of years.

561
00:22:20.035 --> 00:22:23.260
And so I think there's something to be said for that.

562
00:22:23.260 --> 00:22:25.100
What did you think about, Sam

563
00:22:25.100 --> 00:22:29.090
on her comments about implementation the challenges?

564
00:22:29.090 --> 00:22:30.090
Yeah, exactly.

565
00:22:30.090 --> 00:22:31.530
I thought it was interesting that

566
00:22:31.530 --> 00:22:33.030
they've not changed what they're doing

567
00:22:33.030 --> 00:22:34.960
in terms of AI projects.

568
00:22:34.960 --> 00:22:37.410
Because of some of the fact that it's hard

569
00:22:37.410 --> 00:22:39.682
I think they recognize-
They've doubled down on it.

570
00:22:39.682 --> 00:22:41.890
They've doubled down on it.

571
00:22:41.890 --> 00:22:44.580
That in itself is a little bit of a validation

572
00:22:44.580 --> 00:22:46.160
that what they're doing they believe in.

573
00:22:46.160 --> 00:22:48.973
That even though it's hard they're still continuing.

574
00:22:50.550 --> 00:22:51.850
I think they're picking projects

575
00:22:51.850 --> 00:22:53.290
based on their strategic focus

576
00:22:53.290 --> 00:22:55.880
rather than picking a project to do AI.

577
00:22:55.880 --> 00:22:58.560
That says something that just keeps coming up.

578
00:22:58.560 --> 00:23:02.390
No one's out here saying, "Hey, let's do some AI today.

579
00:23:02.390 --> 00:23:05.040
It's Monday morning, let's do some AI."

580
00:23:05.040 --> 00:23:06.790
These people are trying to deliver

581
00:23:06.790 --> 00:23:10.170
on things that are in their organization's strategy.

582
00:23:10.170 --> 00:23:13.750
And it turns out that in many cases, AI is the tool.

583
00:23:13.750 --> 00:23:16.850
But if it's not, that's okay too. But it often is.

584
00:23:16.850 --> 00:23:20.800
Strategy with AI, not strategy for AI.

585
00:23:20.800 --> 00:23:21.810
Where did I read that?

586
00:23:21.810 --> 00:23:23.948
Yeah, I know, right? Didn't you write it?

587
00:23:23.948 --> 00:23:26.140
(chuckles) That was all David.

588
00:23:26.140 --> 00:23:28.960
The other thing I thought was quite insightful

589
00:23:28.960 --> 00:23:32.520
and she went there very early on is about

590
00:23:33.420 --> 00:23:37.940
trust and ethical use of AI.

591
00:23:37.940 --> 00:23:40.980
And sort of reigning in the AI solution

592
00:23:40.980 --> 00:23:42.100
to make sure it's ethical.

593
00:23:42.100 --> 00:23:46.540
But also using it to uncover hidden bias.

594
00:23:46.540 --> 00:23:48.400
So that you could be more inclusive,

595
00:23:48.400 --> 00:23:51.560
you could be more aware of the workforce

596
00:23:51.560 --> 00:23:54.050
of your customers, of the ecosystem.

597
00:23:54.050 --> 00:23:56.510
What I liked was particularly insightful there

598
00:23:56.510 --> 00:24:01.510
was the need not to view all data as equally neutral.

599
00:24:02.400 --> 00:24:05.890
And that depending on where the data goes to

600
00:24:05.890 --> 00:24:07.800
very harmful things can come out of it.

601
00:24:07.800 --> 00:24:11.020
Which necessitates another layer.

602
00:24:11.020 --> 00:24:11.853
She hinted on it, right?

603
00:24:11.853 --> 00:24:15.890
There's a layer of awareness and governance.

604
00:24:15.890 --> 00:24:20.800
And a mix of tech artifacts as well as process protocols

605
00:24:20.800 --> 00:24:25.800
to make sure that the outputs are ethical and unbiased.

606
00:24:26.260 --> 00:24:29.920
And she talked a lot about the need for mindset change

607
00:24:29.920 --> 00:24:30.980
for that to work.

608
00:24:30.980 --> 00:24:33.950
That people have to ask, what am I using it for?

609
00:24:33.950 --> 00:24:34.950
What is it gonna do?

610
00:24:34.950 --> 00:24:36.100
What could go wrong?

611
00:24:36.100 --> 00:24:39.510
And that constantly probing

612
00:24:39.510 --> 00:24:44.400
for things that can go wrong so that you could preempt it.

613
00:24:44.400 --> 00:24:47.400
But also she saw it as a tool to prevent bias.

614
00:24:47.400 --> 00:24:49.047
We've seen so many stories out there about,

615
00:24:49.047 --> 00:24:50.930
"Oh, this AI system is biased.

616
00:24:50.930 --> 00:24:52.487
That AI system has caused bias."

617
00:24:52.487 --> 00:24:54.450
And I'm not denying those are true.

618
00:24:54.450 --> 00:24:56.668
I'm not a biased denier.

619
00:24:56.668 --> 00:24:57.501
(Trevor chuckles)

620
00:24:57.501 --> 00:25:00.294
There's an opportunity that she saw in that data

621
00:25:00.294 --> 00:25:01.940
to rectify that.

622
00:25:01.940 --> 00:25:02.973
They are a fact.

623
00:25:04.890 --> 00:25:08.490
No amount of retrofitting is gonna take that away.

624
00:25:08.490 --> 00:25:10.840
But she also said we don't ignore the data either.

625
00:25:10.840 --> 00:25:13.810
We just have to know what that provenance is.

626
00:25:13.810 --> 00:25:16.320
I think that's a great point, Sam.

627
00:25:16.320 --> 00:25:17.690
Thanks for joining us today.

628
00:25:17.690 --> 00:25:20.240
Next time Shervin and I will talk with Chris Couch.

629
00:25:20.240 --> 00:25:21.760
Chris is the senior vice president

630
00:25:21.760 --> 00:25:24.540
and chief technology officer at Cooper Standard.

631
00:25:24.540 --> 00:25:25.373
Please join us.

632
00:25:26.980 --> 00:25:29.740
Thanks for listening to "Me, Myself and AI."

633
00:25:29.740 --> 00:25:31.360
If you're enjoying the show

634
00:25:31.360 --> 00:25:33.470
take a minute to write us a review.

635
00:25:33.470 --> 00:25:35.090
If you send us a screenshot

636
00:25:35.090 --> 00:25:36.310
we'll send you a collection

637
00:25:36.310 --> 00:25:40.010
of MIT SMRs best articles on artificial intelligence

638
00:25:40.010 --> 00:25:41.910
free for a limited time.

639
00:25:41.910 --> 00:25:46.733
Send your review screenshot to smrfeedback@mit.edu.

640
00:25:46.733 --> 00:25:49.316
(gentle music)