WEBVTT

1
00:00:01.630 --> 00:00:03.830
In our previous video Roz Picard

2
00:00:03.830 --> 00:00:07.210
from the MIT Media Lab talks about affective computing,

3
00:00:07.210 --> 00:00:09.660
which lets researchers gain valuable insights

4
00:00:09.660 --> 00:00:11.550
into our emotions.

5
00:00:11.550 --> 00:00:14.130
Affective computing opens up new opportunities

6
00:00:14.130 --> 00:00:16.610
for organizations to improve team performance,

7
00:00:16.610 --> 00:00:18.590
and employees' well-being.

8
00:00:18.590 --> 00:00:21.280
But it also presents some pitfalls.

9
00:00:21.280 --> 00:00:22.950 line:15% 
People often come to us and they think,

10
00:00:22.950 --> 00:00:26.760 line:15% 
Oh I want to understand when my employees are stressed.

11
00:00:26.760 --> 00:00:29.930 line:15% 
I want to help them manage their stress.

12
00:00:29.930 --> 00:00:34.260
And I want to help them be more productive, and healthy,

13
00:00:34.260 --> 00:00:36.760
and these are all good goals.

14
00:00:36.760 --> 00:00:38.730
This can backfire.

15
00:00:38.730 --> 00:00:41.980
The first thing I would do is have a conversation

16
00:00:41.980 --> 00:00:45.620
with your employees, and each other,

17
00:00:45.620 --> 00:00:49.960
about what you would feel comfortable having measured,

18
00:00:49.960 --> 00:00:52.710
and what you want done with that data.

19
00:00:52.710 --> 00:00:55.890
I find, repeatedly with the groups we work with,

20
00:00:55.890 --> 00:00:57.980
that people want to know,

21
00:00:57.980 --> 00:01:00.050
but they don't want their boss to know.

22
00:01:00.050 --> 00:01:02.740
So I think it's very important that you put in place

23
00:01:02.740 --> 00:01:06.560
things that give insight into themselves,

24
00:01:06.560 --> 00:01:10.090
without them feeling threatened, judged,

25
00:01:10.090 --> 00:01:13.010
observed, evaluated by others.

26
00:01:13.010 --> 00:01:15.364
That's number one in the workplace.

27
00:01:17.870 --> 00:01:19.800
Once your teams trust you,

28
00:01:19.800 --> 00:01:21.480
you can use affective computing

29
00:01:21.480 --> 00:01:24.460
to make some surprising discoveries.

30
00:01:24.460 --> 00:01:27.370
We've done a lot of work in call centers.

31
00:01:27.370 --> 00:01:28.330
There are a lot of ideas.

32
00:01:28.330 --> 00:01:30.410
They're all centered around helping

33
00:01:30.410 --> 00:01:32.410
the call center worker to feel better,

34
00:01:32.410 --> 00:01:34.380
so they can do their job better.

35
00:01:34.380 --> 00:01:37.300
For example if they're having a really high stress day,

36
00:01:37.300 --> 00:01:39.140
why not notice a priori,

37
00:01:39.140 --> 00:01:43.170
if the next call in the cue is a super nasty one,

38
00:01:43.170 --> 00:01:45.030
or a more pleasant one,

39
00:01:45.030 --> 00:01:47.760
and try to load balance those a little bit.

40
00:01:47.760 --> 00:01:49.200
So that the person having a bad day

41
00:01:49.200 --> 00:01:51.820
doesn't get all the bad calls.

42
00:01:51.820 --> 00:01:54.480
Affective computing can also help organizations

43
00:01:54.480 --> 00:01:57.310
improve how people work together in teams.

44
00:01:57.310 --> 00:02:00.550
In one study simple nudges driven by AI,

45
00:02:00.550 --> 00:02:04.840
improved communication and morale in a hard working team.

46
00:02:04.840 --> 00:02:07.580
There was a chat bot that gave you these little nudges,

47
00:02:07.580 --> 00:02:08.940
within a team, you know,

48
00:02:08.940 --> 00:02:11.810
hey, would you be willing to share a little positivity

49
00:02:11.810 --> 00:02:13.200
with somebody on your team, right?

50
00:02:13.200 --> 00:02:14.870
But it never told you what to say,

51
00:02:14.870 --> 00:02:16.080
or that you had to do anything.

52
00:02:16.080 --> 00:02:20.210
Not only did they show the hypothesized improvements

53
00:02:20.210 --> 00:02:22.990
related to their self-esteem, and their mental health,

54
00:02:22.990 --> 00:02:25.080
but afterwards, when they were being interviewed,

55
00:02:25.080 --> 00:02:26.570
like, you know, hey how can we do this better?

56
00:02:26.570 --> 00:02:27.810
And what parts of it were bad?

57
00:02:27.810 --> 00:02:29.150
And what parts of it were good?

58
00:02:29.150 --> 00:02:30.530
One of the things they went off and they said,

59
00:02:30.530 --> 00:02:35.370
is, "You know, we now feel closer to our teammates."

60
00:02:35.370 --> 00:02:37.630
They were having these revelations,

61
00:02:37.630 --> 00:02:39.600
about feeling related things,

62
00:02:39.600 --> 00:02:41.360
that usually like, just don't come out

63
00:02:41.360 --> 00:02:43.490
in the face to face team dynamics for some groups.

64
00:02:43.490 --> 00:02:45.830
But when they gave these little nudges,

65
00:02:45.830 --> 00:02:49.750
that were still quite appropriate to the workplace

66
00:02:49.750 --> 00:02:52.290
they saw underneath some things

67
00:02:52.290 --> 00:02:55.520
that they said made them connect better to each other.

68
00:02:55.520 --> 00:02:57.460
And it was done very efficiently.

69
00:02:57.460 --> 00:02:58.410
Like you don't have to go off

70
00:02:58.410 --> 00:03:01.420
on some big expensive weekend retreat, right?

71
00:03:01.420 --> 00:03:05.280
You can just add these little thoughtful reflection

72
00:03:05.280 --> 00:03:08.510
moments to the daily dynamic.

73
00:03:08.510 --> 00:03:13.510
That one-liner can be such a huge improver

74
00:03:13.960 --> 00:03:16.210
of a team member's mood,

75
00:03:16.210 --> 00:03:20.180
that what you may not realize, is that one statement,

76
00:03:20.180 --> 00:03:21.750
and that positive mood,

77
00:03:21.750 --> 00:03:25.720
can lead their entire brain to think differently.

78
00:03:25.720 --> 00:03:29.710
They will solve creative problems with higher success rates.

79
00:03:29.710 --> 00:03:34.710
It's amazing how a simple, little, sincere compliment,

80
00:03:36.200 --> 00:03:38.970
can impact not just the mood,

81
00:03:38.970 --> 00:03:41.141
but all of the thinking that follows.

82
00:03:43.820 --> 00:03:46.160
Some companies are already taking advantage

83
00:03:46.160 --> 00:03:48.130
of this new technology.

84
00:03:48.130 --> 00:03:49.430
There are a lot of different groups

85
00:03:49.430 --> 00:03:51.770
interested in affective computing these days.

86
00:03:51.770 --> 00:03:54.730
The initial focus was people

87
00:03:54.730 --> 00:03:58.010
who already pay a lot of attention to customers' emotion.

88
00:03:58.010 --> 00:03:59.180
People in marketing.

89
00:03:59.180 --> 00:04:00.880
People in customer service.

90
00:04:00.880 --> 00:04:03.750
People in education, who want to see if their learners

91
00:04:03.750 --> 00:04:06.920
are engaged, disengaged, confused.

92
00:04:06.920 --> 00:04:09.580
Just knowing if the person in front of you

93
00:04:09.580 --> 00:04:11.190
is even paying attention.

94
00:04:11.190 --> 00:04:13.220
If they're even engaging with you.

95
00:04:13.220 --> 00:04:15.470
If they're not already engaging with you,

96
00:04:15.470 --> 00:04:17.030
then everything else you're trying to do

97
00:04:17.030 --> 00:04:18.860
with them is wasted time.

98
00:04:18.860 --> 00:04:21.890
We can also understand their affect

99
00:04:21.890 --> 00:04:25.130
through things that they're touching and doing physically.

100
00:04:25.130 --> 00:04:27.293
Their posture, their gestures,

101
00:04:28.130 --> 00:04:30.150
if they're grabbing a steering wheel, you know,

102
00:04:30.150 --> 00:04:32.010
and how they're holding it.

103
00:04:32.010 --> 00:04:33.570
The dynamics of their movement,

104
00:04:33.570 --> 00:04:36.220
if they're moving in a happy way.

105
00:04:36.220 --> 00:04:39.495
Or if they're moving in sort of a tense angular way.

106
00:04:44.810 --> 00:04:46.010
By measuring emotion

107
00:04:46.010 --> 00:04:48.140
using biometric indicators,

108
00:04:48.140 --> 00:04:50.160
affective computing removes biases,

109
00:04:50.160 --> 00:04:51.870
that can happen when people describe

110
00:04:51.870 --> 00:04:54.000
their own emotional state.

111
00:04:54.000 --> 00:04:57.480
In some cases computers know how you feel about something,

112
00:04:57.480 --> 00:05:00.410
like a product, better than you do.

113
00:05:00.410 --> 00:05:03.020
So if computers know how you're feeling,

114
00:05:03.020 --> 00:05:05.850
can they show empathy for what you're feeling?

115
00:05:05.850 --> 00:05:07.720
Roz is already working on that.

116
00:05:07.720 --> 00:05:10.580
The first time one of my students came to me and said,

117
00:05:10.580 --> 00:05:11.997
you know, "I want to build a computer

118
00:05:11.997 --> 00:05:13.810
that empathizes with people,"

119
00:05:13.810 --> 00:05:15.270
I tried to talk him out of it.

120
00:05:15.270 --> 00:05:17.940
I thought like, this is not going to work.

121
00:05:17.940 --> 00:05:20.500
And he told me about this terrible flight

122
00:05:20.500 --> 00:05:24.040
he had where the customer handling was awful,

123
00:05:24.040 --> 00:05:26.500
and at the end of it he was determined,

124
00:05:26.500 --> 00:05:27.620
I'm going to build a computer

125
00:05:27.620 --> 00:05:28.880
that handles people's feelings

126
00:05:28.880 --> 00:05:31.750
better than customer service people did.

127
00:05:31.750 --> 00:05:34.100
And I thought, wow this is going to be hard.

128
00:05:34.100 --> 00:05:39.100
We built a computer dialog that empathized with a person,

129
00:05:39.620 --> 00:05:42.070
when they encountered a frustrating experience.

130
00:05:42.070 --> 00:05:45.560
It used active listening, and empathy, and sympathy.

131
00:05:45.560 --> 00:05:48.160
It did not pretend to be a person.

132
00:05:48.160 --> 00:05:50.860
It just said, you know, this computer apologizes,

133
00:05:50.860 --> 00:05:52.740
and tried to be very computer like.

134
00:05:52.740 --> 00:05:54.390
We did a randomized controlled trial,

135
00:05:54.390 --> 00:05:57.020
where we compared it to another chat,

136
00:05:57.020 --> 00:05:59.140
that was just kind of friendly, and chatty,

137
00:05:59.140 --> 00:06:02.360
and that just asked the individual

138
00:06:02.360 --> 00:06:04.410
to vent about their situation.

139
00:06:04.410 --> 00:06:07.040
We found that the empathetic one

140
00:06:07.040 --> 00:06:11.360
was significantly better at reducing frustration.

141
00:06:11.360 --> 00:06:13.360
And afterwards, we were like,

142
00:06:13.360 --> 00:06:16.660
wow I didn't think a computer could pull this off.

143
00:06:16.660 --> 00:06:18.500
And we were doing this with smart people, right?

144
00:06:18.500 --> 00:06:21.000
Who knew that the computers had no feelings.

145
00:06:21.000 --> 00:06:23.350
Now there are some caveats with that.

146
00:06:23.350 --> 00:06:28.340
What we could do was pull off scripted empathy.

147
00:06:28.340 --> 00:06:31.040
Empathy that was very tightly scripted

148
00:06:31.040 --> 00:06:33.810
to say the right thing, in the right situation.

149
00:06:33.810 --> 00:06:36.800
However computers today have to be limited

150
00:06:36.800 --> 00:06:39.440
to very narrow situations before they can do that.

151
00:06:39.440 --> 00:06:41.690
If you put them in a general situation,

152
00:06:41.690 --> 00:06:43.963
to handle it, they're not ready yet.

153
00:06:48.820 --> 00:06:51.540
Affective computing, powered by AI,

154
00:06:51.540 --> 00:06:54.260
is opening up whole new opportunities,

155
00:06:54.260 --> 00:06:56.310
including improving team performance

156
00:06:56.310 --> 00:06:58.410
and employees' well-being.

157
00:06:58.410 --> 00:07:01.080
The ability to learn together makes a huge difference

158
00:07:01.080 --> 00:07:02.870
in how we, and the companies we work for,

159
00:07:02.870 --> 00:07:06.690
can operate more effectively and productively.

160
00:07:06.690 --> 00:07:09.110
In our next video Cesar Hidalgo

161
00:07:09.110 --> 00:07:12.203
takes a closer look at this topic of collective learning.