1
00:00:10,143 --> 00:00:14,244
So some of the smallest
most insignificant words

2
00:00:14,244 --> 00:00:17,995
we use everyday, can reflect
a lot about who we are.

3
00:00:17,995 --> 00:00:20,137
And I say this not as a linguist,

4
00:00:20,137 --> 00:00:23,506
or a computer scientist
but as a social psychologist.

5
00:00:23,506 --> 00:00:25,743
And today I'd like to tell you a story

6
00:00:25,743 --> 00:00:28,345
that summarizes a lot of the research

7
00:00:28,345 --> 00:00:31,060
that my colleagues,
my students and I have done,

8
00:00:31,060 --> 00:00:34,623
that have helped me
to come to this realization.

9
00:00:34,623 --> 00:00:38,137
Now several years ago I was studying
the nature of traumatic experience

10
00:00:38,137 --> 00:00:40,607
and how it is related to physical health

11
00:00:40,607 --> 00:00:44,927
and what I kept finding,
that just completely perplexed me.

12
00:00:44,927 --> 00:00:48,965
Basically when people have a major
traumatic experience in their life,

13
00:00:48,965 --> 00:00:52,355
they are much more likely
to get sick after that event,

14
00:00:52,355 --> 00:00:54,462
if they keep the events secret,

15
00:00:54,468 --> 00:01:02,145
than if they actually
talk to other people.

16
00:01:02,145 --> 00:01:07,127
So, this really bugged me.
So keeping a secret it seems

17
00:01:07,127 --> 00:01:09,319
is somehow toxic.

18
00:01:09,319 --> 00:01:12,320
So this led me to run some experiments

19
00:01:12,320 --> 00:01:14,442
where we brought people in the laboratory

20
00:01:14,442 --> 00:01:16,863
and we asked them to write about

21
00:01:16,863 --> 00:01:19,119
the most traumatic
experiences they've had,

22
00:01:19,119 --> 00:01:21,375
especially if they'd kept them secret.

23
00:01:21,375 --> 00:01:23,995
And these were big traumas,
these were things like rape.

24
00:01:24,019 --> 00:01:27,808
They were like
major public humiliations or failure.

25
00:01:27,808 --> 00:01:31,285
And the results that we got
from this study were stunning.

26
00:01:31,285 --> 00:01:35,725
We discovered that having people write
as little as fifteen minutes a day,

27
00:01:35,725 --> 00:01:37,740
for 3 or 4 consecutive days,

28
00:01:37,740 --> 00:01:41,479
brought about meaningful changes
in people's physical health

29
00:01:41,479 --> 00:01:44,015
and even their immune function.

30
00:01:44,015 --> 00:01:51,006
Translating up, saying experiences
into words makes a difference, but why?

31
00:01:51,006 --> 00:01:53,465
Since then there have been
hundreds of studies

32
00:01:53,465 --> 00:01:55,942
done by labs all over the world
trying to answer this

33
00:01:55,966 --> 00:01:59,764
and they haven't come up
with a single explanation.

34
00:01:59,764 --> 00:02:03,536
My own approach was to
actually look at the essays

35
00:02:03,536 --> 00:02:06,125
that these people wrote,
and try to figure out,

36
00:02:06,125 --> 00:02:08,856
was there something about
the essays that could predict

37
00:02:08,856 --> 00:02:11,565
who would benefit
from writing versus who wouldn't?

38
00:02:11,565 --> 00:02:14,013
I tried and I couldn't figure it out.

39
00:02:14,013 --> 00:02:16,745
So I got a number of psychologists
and other experts

40
00:02:16,745 --> 00:02:19,689
to read and write hundreds
of these essays,

41
00:02:19,689 --> 00:02:22,974
and they couldn't see a pattern either;

42
00:02:22,974 --> 00:02:25,053
I needed to try some other strategies.

43
00:02:25,053 --> 00:02:27,871
So, with the help of
one of my graduate students,

44
00:02:27,871 --> 00:02:30,228
Martha Francis,
we wrote a computer program.

45
00:02:30,228 --> 00:02:33,900
And the idea of this computer program
was to go into any given text

46
00:02:33,900 --> 00:02:36,380
and calculate the percentage
of words in that text

47
00:02:36,380 --> 00:02:39,866
that were positive emotion words,
negative emotion words

48
00:02:39,866 --> 00:02:42,004
or words related to topics such as

49
00:02:42,004 --> 00:02:46,082
death or sex or violence
or religion or family.

50
00:02:46,082 --> 00:02:49,190
And as long as we were writing
the computer program,

51
00:02:49,190 --> 00:02:52,190
I thought well let's go ahead and
throw in some parts of speech,

52
00:02:52,214 --> 00:02:56,715
pronouns, prepositions. Why?
Because it was easy, who cares?

53
00:02:56,715 --> 00:03:01,413
So, I go back, start to analyze
these traumatic essays,

54
00:03:01,413 --> 00:03:06,013
and quickly discover that the content
of what people were writing about

55
00:03:06,013 --> 00:03:09,243
didn't matter in terms of
if they improved in their health or not.

56
00:03:09,243 --> 00:03:13,683
Instead, it was these junk words,
pronouns, and articles,

57
00:03:13,683 --> 00:03:17,573
and prepositions and so forth,
that did matter.

58
00:03:17,578 --> 00:03:19,388
Now think about this.

59
00:03:19,388 --> 00:03:23,120
Here people are writing about
deeply disturbing issues,

60
00:03:23,120 --> 00:03:29,034
and the actual topics that dealt with
tragedies, devastation, horrible things,

61
00:03:29,034 --> 00:03:31,512
the topics themselves
and the words associated

62
00:03:31,512 --> 00:03:33,890
with those topics made no difference.

63
00:03:33,890 --> 00:03:38,622
Instead these little words
like "I" and "the" and "and" did matter.

64
00:03:38,622 --> 00:03:41,889
I'd been looking for the obvious,
but in fact

65
00:03:41,889 --> 00:03:44,758
I'd been paying attention
to what people were saying,

66
00:03:44,758 --> 00:03:47,637
but not how they were saying it.

67
00:03:47,637 --> 00:03:51,883
So how do I go about
analyzing "what" versus "how"?

68
00:03:51,883 --> 00:03:55,026
Well, it turns out that they're
different kinds of classes of words

69
00:03:55,050 --> 00:03:57,757
that look at this distinction,
and one of them is

70
00:03:57,757 --> 00:04:00,231
if you're looking at
what people are writing about,

71
00:04:00,231 --> 00:04:02,278
you look at what are called content words.

72
00:04:02,278 --> 00:04:06,582
These are nouns and regular verbs
and adjectives and some adverbs.

73
00:04:06,582 --> 00:04:09,871
These are the stuff of thought,
these were the stuff of communication.

74
00:04:09,871 --> 00:04:11,976
We were trying to talk to somebody.

75
00:04:11,976 --> 00:04:16,543
Google and search terms
are all based on these content words.

76
00:04:16,543 --> 00:04:19,165
The other group of words
are a class of words

77
00:04:19,165 --> 00:04:21,526
that are generally called function words.

78
00:04:21,526 --> 00:04:27,084
And function words are made up
of the most boring words you can imagine.

79
00:04:27,084 --> 00:04:31,141
They're made up of pronouns:
"I", "me", "he", "she";

80
00:04:31,141 --> 00:04:37,560
prepositions: "to", "of", "for";
auxiliary verbs: "am", "is", "have".

81
00:04:37,560 --> 00:04:42,128
I'll have to wake you up if I keep talking
about these function words.

82
00:04:42,128 --> 00:04:46,758
But it turns out these
function words are really interesting,

83
00:04:46,758 --> 00:04:52,461
because, first of all, there's only about
five hundred function words in English,

84
00:04:52,461 --> 00:04:54,535
so they account for far less than 1%

85
00:04:54,535 --> 00:04:57,349
of all the words we know,
we hear, we read.

86
00:04:57,349 --> 00:05:00,494
Nevertheless, they reflect 55% to 60%

87
00:05:00,494 --> 00:05:03,639
of all the words that
we are surrounded with,

88
00:05:03,639 --> 00:05:06,784
they're everywhere,
but we don't pay attention to them.

89
00:05:06,784 --> 00:05:10,165
In English and in other languages,
they're the shortest words there are,

90
00:05:10,189 --> 00:05:12,710
and when they're spoken
or when you're reading,

91
00:05:12,710 --> 00:05:16,583
they zip into your brain
at the speed of less than 0.2 seconds,

92
00:05:16,583 --> 00:05:20,538
meaning that they're processed
essentially non-consciously.

93
00:05:20,538 --> 00:05:23,110
But there's something
even more interesting about them,

94
00:05:23,134 --> 00:05:26,897
they are social,
they are profoundly social.

95
00:05:26,897 --> 00:05:29,612
Let me give me an example,
let's say you're walking along,

96
00:05:29,636 --> 00:05:32,396
you see a note on the ground,
you pick it up and it says,

97
00:05:32,396 --> 00:05:35,583
"I am placing it on the table."

98
00:05:35,583 --> 00:05:37,732
Well, that kinda makes sense,
kinda doesn't.

99
00:05:37,732 --> 00:05:40,304
"I'm placing it on the table".
There's 2 content words:

100
00:05:40,328 --> 00:05:41,856
"placing" and "table";

101
00:05:41,856 --> 00:05:47,955
all the rest are function words:
"I", "am", "it", "on", "the".

102
00:05:47,955 --> 00:05:51,328
Now the reason this doesn't
make sense to most of us is

103
00:05:51,328 --> 00:05:53,444
who was "I"? No idea.

104
00:05:53,444 --> 00:05:56,687
"Am" implies present tense.
When was it written?

105
00:05:56,692 --> 00:05:59,221
"It"? Pfft, no idea what "it" is.

106
00:05:59,221 --> 00:06:02,733
"On the table", "the table"
means it's a table

107
00:06:02,733 --> 00:06:04,832
that the author knew about

108
00:06:04,838 --> 00:06:07,704
and the intended recipient
of this note knew about,

109
00:06:07,704 --> 00:06:09,373
but nobody else did.

110
00:06:09,373 --> 00:06:13,752
And, in fact, this note only
has meaning to the author

111
00:06:13,752 --> 00:06:15,833
and the recipient of the note

112
00:06:15,833 --> 00:06:18,994
at a particular time,
in a particular location.

113
00:06:19,008 --> 00:06:22,057
And, in fact, if I took
that note to this author now

114
00:06:22,057 --> 00:06:24,458
6 months later and say,
"What's this all about?"

115
00:06:24,458 --> 00:06:26,649
There's a good chance
that the author will say,

116
00:06:26,673 --> 00:06:28,716
"No idea."

117
00:06:30,342 --> 00:06:32,960
Function words are social,

118
00:06:32,960 --> 00:06:34,567
they tell us about the author,

119
00:06:34,567 --> 00:06:36,234
they tell us about the relationship

120
00:06:36,234 --> 00:06:37,949
between the author and the recipient

121
00:06:37,973 --> 00:06:41,578
and the relationship between
the author and the topic itself.

122
00:06:41,578 --> 00:06:45,731
And this is the heart of what
I want to talk to you about today.

123
00:06:45,731 --> 00:06:49,535
By analyzing function words,
we start to get a sense

124
00:06:49,535 --> 00:06:51,945
of who people are,
what their relationships are,

125
00:06:51,945 --> 00:06:55,811
how they think about themselves
and how they connect with others.

126
00:06:57,451 --> 00:06:59,343
Yeah, there're a lot of function words,

127
00:06:59,343 --> 00:07:02,959
and honestly, I could talk for
several hours about function words.

128
00:07:02,959 --> 00:07:06,634
But I'm going to spare you that
and just focus on a couple today,

129
00:07:06,634 --> 00:07:10,745
to just give you a flavor
of why they're so interesting.

130
00:07:10,745 --> 00:07:12,372
Let's start off with pronouns,

131
00:07:12,372 --> 00:07:14,519
and let's start off
with third-person pronouns

132
00:07:14,519 --> 00:07:16,287
like "he", "she", "they".

133
00:07:17,337 --> 00:07:19,718
Now it turns out some people
out there in the world

134
00:07:19,742 --> 00:07:22,543
use these third-person pronouns
at high rates

135
00:07:22,543 --> 00:07:24,947
and other people at low rates.

136
00:07:24,947 --> 00:07:27,534
What kind of person would use them?

137
00:07:27,534 --> 00:07:31,232
Well, you have to think about pronouns
and all function words

138
00:07:31,232 --> 00:07:35,784
in terms of where are
people paying attention.

139
00:07:35,784 --> 00:07:37,880
If you are using
these third-person pronouns,

140
00:07:37,904 --> 00:07:40,527
by definition you're paying
attention to other people.

141
00:07:40,527 --> 00:07:43,571
You care about other people,
you're thinking about other people,

142
00:07:43,571 --> 00:07:45,579
and people who use these at high rates

143
00:07:45,579 --> 00:07:47,709
are much more socially engaged.

144
00:07:47,709 --> 00:07:49,981
We can analyze emails, tweets and so forth

145
00:07:49,981 --> 00:07:52,449
and get a sense of someone's
social engagement

146
00:07:52,449 --> 00:07:54,097
just by looking at this.

147
00:07:54,097 --> 00:07:57,815
How about first-person singular pronouns,
"I", "me" and "my"?

148
00:07:59,335 --> 00:08:01,713
OK, using the attentional arguments,

149
00:08:01,713 --> 00:08:04,629
someone who's attending
to their thoughts, feelings,

150
00:08:04,629 --> 00:08:08,315
behaviors, to themselves in some way
would use these words more.

151
00:08:08,315 --> 00:08:11,615
What kind of person
do you think uses "I" words the most?

152
00:08:11,615 --> 00:08:14,584
I hope, you're sitting there,
you're thinking,

153
00:08:14,584 --> 00:08:17,377
"Well, somebody who's
self-centered, self-important,

154
00:08:17,377 --> 00:08:23,805
narcissistic, hungry for power
and high in status."

155
00:08:24,345 --> 00:08:26,758
You would be completely wrong.

156
00:08:27,278 --> 00:08:30,012
In fact the person who is
highest in status

157
00:08:30,012 --> 00:08:33,196
uses "I" words the least.

158
00:08:33,535 --> 00:08:35,481
Let me rephrase that,

159
00:08:35,481 --> 00:08:39,376
the higher anybody is in status,
the less they use "I" words;

160
00:08:39,376 --> 00:08:44,207
the lower someone is in status,
the higher they use "I" words.

161
00:08:44,207 --> 00:08:47,750
Now, I discovered this
by analyzing emails,

162
00:08:47,750 --> 00:08:50,745
instant messages, natural conversations,

163
00:08:50,745 --> 00:08:52,530
business groups and so forth.

164
00:08:52,530 --> 00:08:54,054
And the effects were huge.

165
00:08:55,024 --> 00:08:57,388
I looked at these results and I thought,

166
00:08:57,388 --> 00:09:00,260
"Wow, this must be true for other people

167
00:09:00,260 --> 00:09:02,472
but it can't possibly be true for me."

168
00:09:03,292 --> 00:09:04,912
You know I love everybody equally.

169
00:09:05,665 --> 00:09:08,398
So I go in and analyze my own emails.

170
00:09:08,398 --> 00:09:10,952
I'm the same as everybody else,

171
00:09:10,952 --> 00:09:15,183
I look at the email that I get
from an undergraduate student,

172
00:09:15,183 --> 00:09:17,136
"Dear Dr Pennebaker, I would like to know

173
00:09:17,160 --> 00:09:18,951
if I could possibly meet with you

174
00:09:18,951 --> 00:09:20,951
because I think I need
to change my grade."

175
00:09:20,975 --> 00:09:25,114
And I write back, "Dear Student,
Thank you so much for your email.

176
00:09:25,114 --> 00:09:29,730
Unfortunately, the way
the grade systems work, blah, blah, blah."

177
00:09:29,730 --> 00:09:32,225
I look at my email to the dean.

178
00:09:32,225 --> 00:09:35,074
"Dear Dean, I'm Jamie Pennebaker
and I would like to ask you

179
00:09:35,074 --> 00:09:38,070
if I could do this and if I could
do that and I could do this."

180
00:09:38,070 --> 00:09:39,915
And the dean writes back, "Dear Jamie,

181
00:09:39,915 --> 00:09:43,959
Thank you so much for your email..."
and so forth.

182
00:09:43,959 --> 00:09:49,469
Now everybody is being completely polite,
nobody's putting anybody down.

183
00:09:49,469 --> 00:09:52,151
This is the language of power in status;

184
00:09:52,151 --> 00:09:54,526
it tells us where people
are paying attention.

185
00:09:54,543 --> 00:09:57,273
A high status person is
looking out at the world,

186
00:09:57,277 --> 00:10:00,532
the low status person tends
to be looking more inwardly.

187
00:10:00,532 --> 00:10:02,247
What about others' states?

188
00:10:02,247 --> 00:10:07,237
Let's move beyond status,
let's look at emotional states.

189
00:10:07,237 --> 00:10:09,088
You would think that someone would be

190
00:10:09,088 --> 00:10:10,755
paying more attention to themselves

191
00:10:10,779 --> 00:10:15,550
if they're in pain. It could be
physical pain or emotional pain.

192
00:10:15,550 --> 00:10:18,568
In fact, if we look at people
who are depressed,

193
00:10:18,568 --> 00:10:20,348
we've done many studies on this,

194
00:10:20,348 --> 00:10:22,368
and we know that people who are depressed

195
00:10:22,368 --> 00:10:23,924
pay attention to themselves more

196
00:10:23,924 --> 00:10:25,900
and they used the word "I" more.

197
00:10:25,900 --> 00:10:28,615
In fact one of our very first studies
looked at the poetry

198
00:10:28,639 --> 00:10:31,008
of suicidal and non-suicidal poets.

199
00:10:31,008 --> 00:10:33,580
Now, we did this research
where we went through,

200
00:10:33,580 --> 00:10:35,162
analyzed their poetry,

201
00:10:35,162 --> 00:10:37,639
and initially I thought,
"Well, the big difference is

202
00:10:37,639 --> 00:10:40,144
in the degree that they use
negative, emotion words."

203
00:10:40,144 --> 00:10:41,852
Not true.

204
00:10:41,852 --> 00:10:43,541
Suicidal and non-suicidal poets

205
00:10:43,541 --> 00:10:46,930
all use negative emotion
words at high rate.

206
00:10:46,930 --> 00:10:50,258
I think it's part of the job description.

207
00:10:50,258 --> 00:10:52,246
(Laughter)

208
00:10:52,246 --> 00:10:54,904
The big difference was
their use of the word "I".

209
00:10:54,904 --> 00:10:57,946
Suicidal poets use the word "I" more.

210
00:10:57,946 --> 00:11:00,271
Consider this poem,
this is by Sylvia Plath

211
00:11:00,271 --> 00:11:03,213
who later committed suicide.

212
00:11:03,213 --> 00:11:06,701
Listen to the way that
she uses the word "I"

213
00:11:06,701 --> 00:11:08,419
and first-person singular.

214
00:11:09,739 --> 00:11:13,348
I'm taking some lines from her poem
"Mad Girl's Love Song".

215
00:11:14,488 --> 00:11:17,451
I shut my eyes and
all the world drops dead;

216
00:11:17,451 --> 00:11:20,995
I lift my lids and all is born again.

217
00:11:20,995 --> 00:11:23,547
(I think I made you up inside my head.)

218
00:11:23,547 --> 00:11:25,575
I fancied you'd return the way you said,

219
00:11:25,575 --> 00:11:28,003
But I grow old and I forget your name.

220
00:11:28,593 --> 00:11:30,773
(I think I made you up inside my head.)

221
00:11:32,633 --> 00:11:35,643
You can almost see Plath

222
00:11:35,643 --> 00:11:39,217
embracing her sorrow,
her misery and so forth

223
00:11:39,217 --> 00:11:44,486
and you can compare her writing
with other poets, non-suicidal poets

224
00:11:44,486 --> 00:11:46,255
who write about lost love.

225
00:11:46,255 --> 00:11:50,012
When they do, you can almost see
them holding it off from a distance,

226
00:11:50,012 --> 00:11:55,189
so they're looking at it from a more
distant, third-person perspective.

227
00:11:55,189 --> 00:11:59,868
Now, there's a really interesting,
important theory within psychology

228
00:11:59,868 --> 00:12:02,946
about depression.
And people who are depressed

229
00:12:02,946 --> 00:12:07,099
are thought to be people who are
very high in self-awareness or self-focus.

230
00:12:07,099 --> 00:12:12,164
And part of this is
they also tend to be extremely honest.

231
00:12:12,164 --> 00:12:19,546
In fact, there are many studies
showing that they have this deficit

232
00:12:19,546 --> 00:12:23,287
and they're not able to have
positive illusions about ourselves.

233
00:12:23,287 --> 00:12:25,573
Those of us who aren't depressed
get by every day

234
00:12:25,597 --> 00:12:28,804
by holding these insane illusions
about the life.

235
00:12:28,804 --> 00:12:32,408
But these people are brutally honest.

236
00:12:32,408 --> 00:12:33,986
Now this made me wonder:

237
00:12:33,986 --> 00:12:36,384
throw away depression for just a second.

238
00:12:36,384 --> 00:12:38,384
Could we turn this entire thing
upside down

239
00:12:38,408 --> 00:12:40,174
and find out if depressed people

240
00:12:40,174 --> 00:12:44,797
or if we could use a computer program
as a linguistic lie-detector.

241
00:12:44,797 --> 00:12:47,535
I mean for anybody.
So in fact we did some studies,

242
00:12:47,535 --> 00:12:49,230
where we brought people in the lab,

243
00:12:49,230 --> 00:12:51,185
we induced them to lie or tell the truth,

244
00:12:51,185 --> 00:12:54,233
we analyze court transcripts
of people who were all found guilty,

245
00:12:54,257 --> 00:12:56,679
half of whom were later exonerated,

246
00:12:56,679 --> 00:12:59,924
and the effects were
really quite impressive.

247
00:12:59,924 --> 00:13:01,999
We did a pretty good job at telling

248
00:13:01,999 --> 00:13:04,333
if someone who was telling
the truth versus lying,

249
00:13:04,357 --> 00:13:07,367
and one of the best words
was the use of the word "I".

250
00:13:07,367 --> 00:13:08,941
People who tell the truth

251
00:13:08,941 --> 00:13:12,695
use the word "I" more,
owning what they're saying.

252
00:13:12,695 --> 00:13:16,609
Liars are tending to hold off,
distancing themselves.

253
00:13:17,759 --> 00:13:21,018
Now, lie-detection and depression,

254
00:13:21,018 --> 00:13:23,907
status, are all some things
that we can look at,

255
00:13:23,907 --> 00:13:26,384
but one of the things that
I'm most interested in now

256
00:13:26,408 --> 00:13:27,729
is looking at groups,

257
00:13:27,729 --> 00:13:29,920
looking at the relationship
between two people.

258
00:13:29,944 --> 00:13:32,445
Can you tell how two people
are getting along

259
00:13:32,445 --> 00:13:36,872
by analyzing the way that they're
using function words with each other?

260
00:13:36,872 --> 00:13:38,312
And the answer is "yes".

261
00:13:38,312 --> 00:13:40,832
We'd look at the percentage
of each class of words

262
00:13:40,832 --> 00:13:44,183
and we come up with the metric
that we call language style matching.

263
00:13:44,183 --> 00:13:48,244
And the more that two people are matching
in their function word use,

264
00:13:48,244 --> 00:13:50,041
the more they're on the same page,

265
00:13:50,041 --> 00:13:53,024
the more they're talking
about something in the same way.

266
00:13:53,024 --> 00:13:54,929
Now one place we started to look at this

267
00:13:54,953 --> 00:13:56,674
was with speed dating.

268
00:13:57,326 --> 00:14:00,771
Now, I should tell you
I love speed dating,

269
00:14:00,771 --> 00:14:02,336
(Laughter)

270
00:14:02,336 --> 00:14:04,866
I would never do it in a million years,

271
00:14:04,876 --> 00:14:05,956
(Laughter)

272
00:14:05,956 --> 00:14:08,195
but I encourage all of you
to go do speed dating

273
00:14:08,219 --> 00:14:09,460
and when you do,

274
00:14:09,460 --> 00:14:10,993
invite a researcher along

275
00:14:10,993 --> 00:14:13,576
because there is
no paradigm that is better.

276
00:14:14,666 --> 00:14:17,320
We've been involved
in speed-dating projects

277
00:14:17,320 --> 00:14:19,427
where people come in

278
00:14:19,427 --> 00:14:21,803
and in their 4 minute date,
we tape-record it,

279
00:14:21,803 --> 00:14:23,741
they know we are,

280
00:14:23,741 --> 00:14:26,819
and then we transcribe the way they talk.

281
00:14:26,819 --> 00:14:29,327
The more they match in their language,

282
00:14:29,327 --> 00:14:31,423
the more likely they are
to go out on a date.

283
00:14:31,447 --> 00:14:33,242
We can predict who will go on a date

284
00:14:33,242 --> 00:14:37,591
at rates slightly better than
the people themselves can.

285
00:14:37,591 --> 00:14:40,231
We've done studies
with young dating couples.

286
00:14:40,231 --> 00:14:43,622
To be in our study, they had to give us

287
00:14:43,622 --> 00:14:46,123
10 days of their instant messages or IM's.

288
00:14:46,723 --> 00:14:49,172
And then what we do
is we analyze their IM's

289
00:14:49,172 --> 00:14:51,621
with this style matching,

290
00:14:51,621 --> 00:14:54,070
and we do much, much better than they do

291
00:14:54,070 --> 00:14:56,499
at predicting if they'll be
together 3 months later.

292
00:14:56,523 --> 00:14:58,362
(Laughter)

293
00:14:58,362 --> 00:15:01,077
The fact is, is these words
are telling us how individuals

294
00:15:01,101 --> 00:15:03,551
and pairs of people are connecting.

295
00:15:03,551 --> 00:15:04,611
What about groups?

296
00:15:04,611 --> 00:15:07,051
Now this is an area
that we're now working at.

297
00:15:07,051 --> 00:15:08,575
We're looking at working groups,

298
00:15:08,599 --> 00:15:10,513
some are groups that we've worked with,

299
00:15:10,513 --> 00:15:12,037
people from the business school,

300
00:15:12,061 --> 00:15:14,495
we'd look at people in
the get-to-know-you groups,

301
00:15:14,527 --> 00:15:16,908
we do educational groups.
And what we're finding is

302
00:15:16,932 --> 00:15:19,836
by looking at a group of,
say, 5 or 6 people,

303
00:15:19,836 --> 00:15:22,598
we can now get a sense
of how productive the group will be,

304
00:15:22,622 --> 00:15:25,258
and also how cohesive the group will be,

305
00:15:25,258 --> 00:15:27,969
simply by looking at the style matching.

306
00:15:27,976 --> 00:15:30,912
Now here's where things are
starting to get interesting:

307
00:15:30,912 --> 00:15:33,451
by tracking a group that's interacting

308
00:15:33,451 --> 00:15:35,770
and say they're all interacting online,

309
00:15:35,770 --> 00:15:39,551
we can have a computer monitoring
how the group is behaving.

310
00:15:39,551 --> 00:15:40,867
Imagine for example,

311
00:15:40,867 --> 00:15:43,544
you are in this group,

312
00:15:43,544 --> 00:15:46,211
and a computer coach comes to your group

313
00:15:46,211 --> 00:15:50,190
every now and then,
and a message comes and says,

314
00:15:50,190 --> 00:15:53,296
"You guys are not paying attention
to one another,

315
00:15:53,296 --> 00:15:57,396
you need to be more attentive to what
the other people are saying," or

316
00:15:57,396 --> 00:16:00,396
"You guys for the last few minutes
have strayed off topic,

317
00:16:00,396 --> 00:16:03,716
try to get back on topic,"
or that loud mouth in the group,

318
00:16:03,716 --> 00:16:07,566
the computer comes in and says,
"John, for the last 5 minutes

319
00:16:07,566 --> 00:16:11,142
you said 50% of the words,
why don't you stand back

320
00:16:11,142 --> 00:16:13,884
and encourage others to talk?" (Laughter)

321
00:16:13,884 --> 00:16:17,199
Well, we have now created
a program that does this,

322
00:16:17,199 --> 00:16:20,602
and we've now tested it out
with hundreds of groups

323
00:16:20,602 --> 00:16:24,703
and we are getting very promising results.

324
00:16:24,703 --> 00:16:27,912
Now, you can start to see
why I'm so excited

325
00:16:27,912 --> 00:16:30,244
about this world of function words,

326
00:16:30,244 --> 00:16:33,665
that we're now taking this
in all these directions,

327
00:16:33,665 --> 00:16:36,249
that I never would've thought about.

328
00:16:36,249 --> 00:16:39,533
We've been looking at it in terms
of looking at historical records.

329
00:16:39,533 --> 00:16:43,876
Can you tell if a particular explorer
committed suicide or was murdered?

330
00:16:43,876 --> 00:16:45,611
We've done a project on that.

331
00:16:45,611 --> 00:16:48,056
Can you look at a company and get a sense

332
00:16:48,056 --> 00:16:51,307
of how their internal
communications are working?

333
00:16:51,307 --> 00:16:53,355
How well they are connecting
with the people

334
00:16:53,355 --> 00:16:55,283
in their company or with their clients.

335
00:16:55,283 --> 00:16:59,752
We look at corporate earnings reports
or the quarterly phone calls

336
00:16:59,752 --> 00:17:03,112
to get a sense of the internal
group dynamics of the company.

337
00:17:03,112 --> 00:17:04,636
We've worked with the government

338
00:17:04,660 --> 00:17:06,688
to try to get a sense of terrorist groups

339
00:17:06,688 --> 00:17:09,813
and if they are likely to behave badly.

340
00:17:09,813 --> 00:17:13,155
We've helped people sort out
their love lives.

341
00:17:14,595 --> 00:17:15,978
You can start to see that

342
00:17:15,978 --> 00:17:19,721
by harnessing the power
of these function words,

343
00:17:19,721 --> 00:17:22,948
we can get a sense
of individuals and groups

344
00:17:22,948 --> 00:17:25,325
and how people are connecting.

345
00:17:26,305 --> 00:17:28,224
Now, what I would urge you to do,

346
00:17:29,185 --> 00:17:31,805
I'd like you to go home tonight

347
00:17:31,805 --> 00:17:34,557
and I want you to start looking
at your emails,

348
00:17:34,557 --> 00:17:37,728
your tweets, your IM's or whatever,

349
00:17:37,728 --> 00:17:41,401
and in doing that,
what I hope you start to see is,

350
00:17:41,401 --> 00:17:43,306
first of all you learn a little bit more

351
00:17:43,306 --> 00:17:45,068
about your relationships with others,

352
00:17:45,092 --> 00:17:46,903
but more than anything,

353
00:17:46,903 --> 00:17:49,411
I hope you'll learn
a little bit about yourself.

354
00:17:49,411 --> 00:17:50,800
Thank you very much.

355
00:17:50,977 --> 00:17:56,459
(Applause)