TED2016
Mariano Sigman: Your words may predict your future mental health
馬理阿諾．西格曼: 你的用詞可能透露你未來的精神健康狀態
Filmed:
Readability: 4.5
3,146,887 views
從你的說話方式及寫出來的文字之中就可以預測出你未來的精神狀態嗎？甚至是病發精神錯亂呢？在這場精彩的演講中，神經學家馬理阿諾．西格曼（Mariano Sigman），透過古希臘書籍及對內省來源的分析，他發現到我們的話語中可以反映出我們的內心世界，並講解他如何用文字地圖的演算法，預測精神分裂症的發展。「未來我們也許可以看到一個全然不同的心理健康模式」，西格曼如是說，「而且是基於一種客觀、量化的方式來自動分析出我們所寫的字、我們所說的話」。
Mariano Sigman - Neuroscientist
In his provocative, mind-bending book "The Secret Life of the Mind," neuroscientist Mariano Sigman reveals his life’s work exploring the inner workings of the human brain. Full bio
In his provocative, mind-bending book "The Secret Life of the Mind," neuroscientist Mariano Sigman reveals his life’s work exploring the inner workings of the human brain. Full bio
Double-click the English transcript below to play the video.
00:13
We have historical records that allow us
to know how the ancient Greeks dressed,
to know how the ancient Greeks dressed,
0
1006
5150
我們有歷史紀錄可循，可以讓我們知道
古希臘人如何穿著、
古希臘人如何穿著、
00:18
how they lived,
1
6180
1254
如何生活、
00:19
how they fought ...
2
7458
1522
如何打仗...
00:21
but how did they think?
3
9004
1524
但他們如何思考呢？
00:23
One natural idea is that the deepest
aspects of human thought --
aspects of human thought --
4
11432
4440
有一個很自然的方法就是，
去探索人類最深層的想法——
去探索人類最深層的想法——
00:27
our ability to imagine,
5
15896
1872
我們的想像力、
00:29
to be conscious,
6
17792
1397
自覺力、
00:31
to dream --
7
19213
1231
夢想力、
00:32
have always been the same.
8
20468
1619
是否是一樣的。
00:34
Another possibility
9
22872
1499
另一種可能是，
00:36
is that the social transformations
that have shaped our culture
that have shaped our culture
10
24395
3723
去探索造就我們文化的社會變革，
00:40
may have also changed
the structural columns of human thought.
the structural columns of human thought.
11
28142
3785
這些變革也許就是
改變人類想法的主要因素。
改變人類想法的主要因素。
00:44
We may all have different
opinions about this.
opinions about this.
12
32911
2524
對這一點，大家或許有不同的看法。
00:47
Actually, it's a long-standing
philosophical debate.
philosophical debate.
13
35459
2717
實際上，這是一個存在已久的哲學辯論。
00:50
But is this question
even amenable to science?
even amenable to science?
14
38644
2727
究竟這個問題是否可以
經由科學來處理？
經由科學來處理？
00:54
Here I'd like to propose
15
42834
2506
我的建議是
00:57
that in the same way we can reconstruct
how the ancient Greek cities looked
how the ancient Greek cities looked
16
45364
4772
如同僅藉由一些磚頭，
我們得以重建希臘古都的外貌，
我們得以重建希臘古都的外貌，
01:02
just based on a few bricks,
17
50160
2388
也可用同樣的方式，
01:04
that the writings of a culture
are the archaeological records,
are the archaeological records,
18
52572
4126
藉由一些文化作品、建築歷史、
01:08
the fossils, of human thought.
19
56722
2143
化石，來了解人類的想法。
01:11
And in fact,
20
59905
1174
而實際上，
01:13
doing some form of psychological analysis
21
61103
2206
因為做了一些
01:15
of some of the most ancient
books of human culture,
books of human culture,
22
63333
3544
人類古老文化書籍的心理分析，
01:18
Julian Jaynes came up in the '70s
with a very wild and radical hypothesis:
with a very wild and radical hypothesis:
23
66901
5955
裘利安．傑尼斯在70年代，
發表了一個相當大膽激進的假說：
發表了一個相當大膽激進的假說：
01:24
that only 3,000 years ago,
24
72880
2413
他說，3000年前的人類，
01:27
humans were what today
we would call schizophrenics.
we would call schizophrenics.
25
75317
4888
是我們現在俗稱的
「精神分裂症患者」。
「精神分裂症患者」。
01:33
And he made this claim
26
81753
1508
他會如此主張的原因是，
01:35
based on the fact that the first
humans described in these books
humans described in these books
27
83285
3301
在世界各地不同的傳統及地方，
01:38
behaved consistently,
28
86610
1904
這些書籍裡面所描述的人類行為
01:40
in different traditions
and in different places of the world,
and in different places of the world,
29
88538
3016
01:43
as if they were hearing and obeying voices
30
91578
3532
似乎不約而同地都會服從
01:47
that they perceived
as coming from the Gods,
as coming from the Gods,
31
95134
3040
他們認為是從神祗那邊傳來的聲音。
01:50
or from the muses ...
32
98198
1198
01:52
what today we would call hallucinations.
33
100063
2769
而如今，我們會稱之為
「幻聽」或「幻覺」。
「幻聽」或「幻覺」。
01:55
And only then, as time went on,
34
103888
2626
隨著時間的洗禮，
01:58
they began to recognize
that they were the creators,
that they were the creators,
35
106538
3651
他們開始認知到
那些聲音是他們自己創造的，
那些聲音是他們自己創造的，
02:02
the owners of these inner voices.
36
110213
2515
他們就是那些內在聲音的主人。
02:05
And with this, they gained introspection:
37
113316
2715
有了這樣的認知，他們學會了自省：
02:08
the ability to think
about their own thoughts.
about their own thoughts.
38
116055
2483
一種反思自己想法的能力。
02:11
So Jaynes's theory is that consciousness,
39
119785
3397
所以傑尼斯對「意識」的理論就是，
02:15
at least in the way we perceive it today,
40
123206
3166
至少現今我們覺察到「意識」、
02:18
where we feel that we are the pilots
of our own existence --
of our own existence --
41
126396
3540
感受到我們自己就是
人生導師的體悟
人生導師的體悟
02:21
is a quite recent cultural development.
42
129960
2737
是相當近代的文化發展。
02:25
And this theory is quite spectacular,
43
133456
1786
這理論相當特別，
02:27
but it has an obvious problem
44
135266
1433
但它有一個很明顯的問題就是，
02:28
which is that it's built on just a few
and very specific examples.
and very specific examples.
45
136723
3992
它是建立在極少又特定的案例上。
02:33
So the question is whether the theory
46
141085
1763
所以問題是，
02:34
that introspection built up in human
history only about 3,000 years ago
history only about 3,000 years ago
47
142872
4751
3000年來人類才建立起
自省能力的這個理論，
自省能力的這個理論，
02:39
can be examined in a quantitative
and objective manner.
and objective manner.
48
147647
2984
是否可以經得起「量化」
且「客觀」的考驗。
且「客觀」的考驗。
02:43
And the problem of how
to go about this is quite obvious.
to go about this is quite obvious.
49
151543
3563
至於要如何做的問題，
也是相當簡單明瞭。
也是相當簡單明瞭。
02:47
It's not like Plato woke up one day
and then he wrote,
and then he wrote,
50
155130
3460
但我的意思並非，比如，
柏拉圖有一天突然醒來說，
柏拉圖有一天突然醒來說，
02:50
"Hello, I'm Plato,
51
158614
1659
「哈囉！我是柏拉圖，
02:52
and as of today, I have
a fully introspective consciousness."
a fully introspective consciousness."
52
160297
2889
我今天，擁有完整的自省意識了」
那樣的簡單而已。
那樣的簡單而已。
02:55
(Laughter)
53
163210
2293
（笑聲）
02:57
And this tells us actually
what is the essence of the problem.
what is the essence of the problem.
54
165527
3333
而這告訴我們，我們要找出
問題的本質為何。
問題的本質為何。
03:01
We need to find the emergence
of a concept that's never said.
of a concept that's never said.
55
169467
4055
我們必須找到從來沒有被
談論過的概念。
談論過的概念。
03:06
The word introspection
does not appear a single time
does not appear a single time
56
174434
4310
「自省」這個字，
在這些書本中從未出現過一次。
03:10
in the books we want to analyze.
57
178768
1919
03:13
So our way to solve this
is to build the space of words.
is to build the space of words.
58
181728
4087
所以為了解決這個問題，
我們要建立一個文字的空間。
我們要建立一個文字的空間。
03:18
This is a huge space
that contains all words
that contains all words
59
186571
3287
在這個大空間裡，
包含了相當多的字，
包含了相當多的字，
03:21
in such a way that the distance
between any two of them
between any two of them
60
189882
2802
用這種方式，可以量測出
03:24
is indicative of how
closely related they are.
closely related they are.
61
192708
2883
兩個字彼此之間的
關聯性程度。
關聯性程度。
03:28
So for instance,
62
196460
1151
舉個例子，
03:29
you want the words "dog" and "cat"
to be very close together,
to be very close together,
63
197635
2897
你會想，「狗」、「貓」
應該是比較有關聯性的，
應該是比較有關聯性的，
03:32
but the words "grapefruit" and "logarithm"
to be very far away.
to be very far away.
64
200556
3831
但「葡萄柚」和「對數」
就沒甚麼關聯了。
就沒甚麼關聯了。
03:36
And this has to be true
for any two words within the space.
for any two words within the space.
65
204809
3896
而在這個空間裡的任何兩個字，
都必須是可以被量測出來的。
都必須是可以被量測出來的。
03:41
And there are different ways
that we can construct the space of words.
that we can construct the space of words.
66
209626
3341
而我們有很多方式
可以建立起這些字的空間架構，
可以建立起這些字的空間架構，
03:44
One is just asking the experts,
67
212991
1643
方法一，是只要請教專家就行了，
03:46
a bit like we do with dictionaries.
68
214658
1896
有點類似查字典。
03:48
Another possibility
69
216896
1428
另一個可行的方法是，
03:50
is following the simple assumption
that when two words are related,
that when two words are related,
70
218348
3715
當兩個字出現關聯性時，
去追蹤它們的預設狀況，
去追蹤它們的預設狀況，
03:54
they tend to appear in the same sentences,
71
222087
2349
它們可能會出現在同一句、
03:56
in the same paragraphs,
72
224460
1453
同一段落、
03:57
in the same documents,
73
225937
1770
或同一文件中，
03:59
more often than would be expected
just by pure chance.
just by pure chance.
74
227731
3182
多於「偶然」地出現。
04:04
And this simple hypothesis,
75
232231
2050
在這個簡單的前提下，
04:06
this simple method,
76
234305
1306
這個單純且帶有運算技巧
的方法必須好用，
的方法必須好用，
04:07
with some computational tricks
77
235635
1607
04:09
that have to do with the fact
78
237266
1389
04:10
that this is a very complex
and high-dimensional space,
and high-dimensional space,
79
238679
3064
而這個複雜且高維度的空間，
04:13
turns out to be quite effective.
80
241767
1665
事後證明，相當有效。
04:16
And just to give you a flavor
of how well this works,
of how well this works,
81
244155
2802
向各位介紹一下，它多有效，
04:18
this is the result we get when
we analyze this for some familiar words.
we analyze this for some familiar words.
82
246981
3912
我們分析了一些經常用到的字，
04:23
And you can see first
83
251607
1185
首先你可以看到，
04:24
that words automatically organize
into semantic neighborhoods.
into semantic neighborhoods.
84
252816
3278
這些詞彙會自動地歸納成
語義相近的相鄰群組，
語義相近的相鄰群組，
04:28
So you get the fruits, the body parts,
85
256118
2217
所以你可看到，水果跟身體部位，
04:30
the computer parts,
the scientific terms and so on.
the scientific terms and so on.
86
258359
2425
電腦與科學字彙等等。
04:33
The algorithm also identifies
that we organize concepts in a hierarchy.
that we organize concepts in a hierarchy.
87
261119
4222
演算法也可以把我們要
整理的概念分門別類出來。
整理的概念分門別類出來。
04:37
So for instance,
88
265852
1151
舉個例子，
04:39
you can see that the scientific terms
break down into two subcategories
break down into two subcategories
89
267027
3597
你可以看到，科學的字彙
被拆解成兩個子類，
被拆解成兩個子類，
04:42
of the astronomic and the physics terms.
90
270648
2100
分別是太空與物理的詞彙。
04:45
And then there are very fine things.
91
273338
2246
然後你會發現一件好玩的事，
04:47
For instance, the word astronomy,
92
275608
1905
舉個例子，「天文學」這個詞彙，
04:49
which seems a bit bizarre where it is,
93
277537
1815
它應該擺的位置
04:51
is actually exactly where it should be,
94
279376
2048
與它現在的位置
04:53
between what it is,
95
281448
1595
好像不太搭嘎，
04:55
an actual science,
96
283067
1270
它現在介於真實科學與
04:56
and between what it describes,
97
284361
1536
天文學之間，偏向科學的位置，
04:57
the astronomical terms.
98
285921
1492
而它自己卻是一個天文學的字彙。
05:00
And we could go on and on with this.
99
288182
1891
我們可以持續尋找
其它類似的情況。
其它類似的情況。
05:02
Actually, if you stare
at this for a while,
at this for a while,
100
290097
2060
實際上，如果你盯著這些字一陣子，
05:04
and you just build random trajectories,
101
292181
1858
然後隨機搭配連結一下這些字，
05:06
you will see that it actually feels
a bit like doing poetry.
a bit like doing poetry.
102
294063
3166
你會覺得好像自己在吟詩。
05:10
And this is because, in a way,
103
298018
1882
那是因為，在某種程度上，
05:11
walking in this space
is like walking in the mind.
is like walking in the mind.
104
299924
2940
在這些空間字彙裡漫遊，
就像是在腦海中吟詩一樣。
就像是在腦海中吟詩一樣。
05:16
And the last thing
105
304027
1617
最後，
05:17
is that this algorithm also identifies
what are our intuitions,
what are our intuitions,
106
305668
4040
演算法也能辨識出人類的直覺字彙，
05:21
of which words should lead
in the neighborhood of introspection.
in the neighborhood of introspection.
107
309732
3896
並歸納到內省的相鄰字彙中。
05:25
So for instance,
108
313652
1223
舉個例子，
05:26
words such as "self," "guilt,"
"reason," "emotion,"
"reason," "emotion,"
109
314899
3979
像是自我、內疚、理由、情緒
05:30
are very close to "introspection,"
110
318902
1889
與內省相關的字彙非常接近，
05:32
but other words,
111
320815
1151
但其它的字，
05:33
such as "red," "football,"
"candle," "banana,"
"candle," "banana,"
112
321990
2167
像是，紅色、足球、蠟燭、香蕉
05:36
are just very far away.
113
324181
1452
就差很遠了。
05:38
And so once we've built the space,
114
326054
2762
所以一旦我們建立起
這樣的詞彙空間，
這樣的詞彙空間，
05:40
the question of the history
of introspection,
of introspection,
115
328840
2826
有關於內省的歷史，
05:43
or of the history of any concept
116
331690
2333
有關與任何概念的歷史，
05:46
which before could seem abstract
and somehow vague,
and somehow vague,
117
334047
4779
以前被認為是抽象或是有點模糊的字彙，
05:50
becomes concrete --
118
338850
1604
都可以變成紮紮實實
05:52
becomes amenable to quantitative science.
119
340478
2738
可以被量化的科學。
05:56
All that we have to do is take the books,
120
344216
2762
而我們要做的就是，
拿起這些書，
拿起這些書，
05:59
we digitize them,
121
347002
1381
把它們數位化，
06:00
and we take this stream
of words as a trajectory
of words as a trajectory
122
348407
2809
然後把這些字，像子彈一樣
06:03
and project them into the space,
123
351240
1969
射到這些字彙空間裡面，
06:05
and then we ask whether this trajectory
spends significant time
spends significant time
124
353233
3754
然後我們問電腦，
這些字彙所行經的軌跡
這些字彙所行經的軌跡
06:09
circling closely to the concept
of introspection.
of introspection.
125
357011
2992
花了多少的時間
才達到內省概念的字彙中。
才達到內省概念的字彙中。
06:12
And with this,
126
360760
1196
有了這些數據，
06:13
we could analyze
the history of introspection
the history of introspection
127
361980
2112
我們就可以分析古希臘傳統中，
06:16
in the ancient Greek tradition,
128
364116
1921
有關於內省的歷史，
06:18
for which we have the best
available written record.
available written record.
129
366061
2602
因為有著最完整的文字記錄。
06:21
So what we did is we took all the books --
130
369631
2255
所以，我們先把這些書，
06:23
we just ordered them by time --
131
371910
2284
按照時間排列，
06:26
for each book we take the words
132
374218
1752
然後把這些字
06:27
and we project them to the space,
133
375994
1961
投射到字彙空間裡面，
06:29
and then we ask for each word
how close it is to introspection,
how close it is to introspection,
134
377979
3032
然後我們問電腦，這些字
與內省有多少的相關性，
與內省有多少的相關性，
06:33
and we just average that.
135
381035
1230
再把它們平均起來。
06:34
And then we ask whether,
as time goes on and on,
as time goes on and on,
136
382590
3198
然後，我們不斷地問電腦問題，
06:37
these books get closer,
and closer and closer
and closer and closer
137
385812
3252
這些書就會
越來越接近內省的概念。
06:41
to the concept of introspection.
138
389088
1754
06:42
And this is exactly what happens
in the ancient Greek tradition.
in the ancient Greek tradition.
139
390866
3801
而這正是古希臘所發生的事。
06:47
So you can see that for the oldest books
in the Homeric tradition,
in the Homeric tradition,
140
395698
3127
各位可以看到在
荷馬時代最古老的書籍，
荷馬時代最古老的書籍，
06:50
there is a small increase with books
getting closer to introspection.
getting closer to introspection.
141
398849
3412
與內省的相關性只有一點點。
06:54
But about four centuries before Christ,
142
402285
2206
但約在西元前400年左右，
06:56
this starts ramping up very rapidly
to an almost five-fold increase
to an almost five-fold increase
143
404515
4708
快速成長了五倍，
07:01
of books getting closer,
and closer and closer
and closer and closer
144
409247
2500
這些書與內省的概念
07:03
to the concept of introspection.
145
411771
1682
越來越接近。
07:06
And one of the nice things about this
146
414159
2424
最棒的是，
07:08
is that now we can ask
147
416607
1198
我們可以問電腦，
07:09
whether this is also true
in a different, independent tradition.
in a different, independent tradition.
148
417829
4147
在不同的、獨立的傳統文化中，
是否也有一樣的現象。
是否也有一樣的現象。
07:14
So we just ran this same analysis
on the Judeo-Christian tradition,
on the Judeo-Christian tradition,
149
422962
3176
所以，我們用同樣的方法，
分析了傳統猶太基督教的書籍，
分析了傳統猶太基督教的書籍，
07:18
and we got virtually the same pattern.
150
426162
2721
也得到了類似的趨勢。
07:21
Again, you see a small increase
for the oldest books in the Old Testament,
for the oldest books in the Old Testament,
151
429548
4635
在最古老的舊約聖經中，
你可以看到它緩慢地增加，
你可以看到它緩慢地增加，
07:26
and then it increases much more rapidly
152
434207
1914
之後在新約聖經中，
它快速地增長，
它快速地增長，
07:28
in the new books of the New Testament.
153
436145
1839
07:30
And then we get the peak of introspection
154
438008
2032
大約西元400年，聖奧古斯丁的《懺悔錄》
內省達到了最高峰。
內省達到了最高峰。
07:32
in "The Confessions of Saint Augustine,"
155
440064
2127
07:34
about four centuries after Christ.
156
442215
1857
07:36
And this was very important,
157
444897
1944
這個方法相當重要，
07:38
because Saint Augustine
had been recognized by scholars,
had been recognized by scholars,
158
446865
3373
因為聖奧古斯丁已經被多位學者、
07:42
philologists, historians,
159
450262
2172
心理學家、歷史學家公認為
07:44
as one of the founders of introspection.
160
452458
2078
是內省的創始人之一。
07:47
Actually, some believe him to be
the father of modern psychology.
the father of modern psychology.
161
455060
3297
有些人認為他是現代心理學之父。
07:51
So our algorithm,
162
459012
1839
所以，我們演算法的優點
07:52
which has the virtue
of being quantitative,
of being quantitative,
163
460875
2842
不僅可以量化、
07:55
of being objective,
164
463741
1263
而且客觀、
07:57
and of course of being extremely fast --
165
465028
2016
當然速度也相當快——
07:59
it just runs in a fraction of a second --
166
467068
2397
幾秒就可以跑完——
08:01
can capture some of the most
important conclusions
important conclusions
167
469489
3503
並捕捉到若使用傳統方法
必須費長時間調查
必須費長時間調查
08:05
of this long tradition of investigation.
168
473016
2222
才能抓到的一些重點。
08:08
And this is in a way
one of the beauties of science,
one of the beauties of science,
169
476317
3651
這也是科學美好的地方，
08:11
which is that now this idea
can be translated
can be translated
170
479992
3476
它可以可以解讀、歸納這想法，
08:15
and generalized to a whole lot
of different domains.
of different domains.
171
483492
2571
然後廣泛應用在許多不同的領域上。
08:18
So in the same way that we asked
about the past of human consciousness,
about the past of human consciousness,
172
486769
4767
或許，最具挑戰性的問題是，
08:23
maybe the most challenging question
we can pose to ourselves
we can pose to ourselves
173
491560
3406
我們用電腦來分析過去的
自我意識發展的方法，
自我意識發展的方法，
08:26
is whether this can tell us something
about the future of our own consciousness.
about the future of our own consciousness.
174
494990
4137
是不是亦可以告訴我們
自我意識的未來趨向呢？
自我意識的未來趨向呢？
08:31
To put it more precisely,
175
499550
1470
更精確地說，
08:33
whether the words we say today
176
501044
2416
我們現在說的話，
08:35
can tell us something
of where our minds will be in a few days,
of where our minds will be in a few days,
177
503484
5197
是否可以告訴我們
接下來的幾天、幾個月或幾年後，
我們的心智會達到什樣的情況。
我們的心智會達到什樣的情況。
08:40
in a few months
178
508705
1151
08:41
or a few years from now.
179
509880
1182
08:43
And in the same way many of us
are now wearing sensors
are now wearing sensors
180
511597
3020
同樣的方式，我們現在很多人
都使用穿戴式偵測器，
都使用穿戴式偵測器，
08:46
that detect our heart rate,
181
514641
1786
可以偵測我們的心跳、
08:48
our respiration,
182
516451
1269
呼吸、
08:49
our genes,
183
517744
1667
基因，
08:51
on the hopes that this may
help us prevent diseases,
help us prevent diseases,
184
519435
3651
讓我們可以預防疾病的發生，
08:55
we can ask whether monitoring
and analyzing the words we speak,
and analyzing the words we speak,
185
523110
3521
我們是否已可以藉由
偵測分析我們所說的話、
偵測分析我們所說的話、
08:58
we tweet, we email, we write,
186
526655
2683
推的文、郵寄的信、寫的文字，
09:01
can tell us ahead of time whether
something may go wrong with our minds.
something may go wrong with our minds.
187
529362
4808
來提前告訴我們，我們的心智
可能要發生問題了？
可能要發生問題了？
09:07
And with Guillermo Cecchi,
188
535087
1534
我跟我的兄弟，吉列爾莫．切基，
09:08
who has been my brother in this adventure,
189
536645
3001
扛起了這項任務。
09:11
we took on this task.
190
539670
1555
09:14
And we did so by analyzing
the recorded speech of 34 young people
the recorded speech of 34 young people
191
542228
5532
我們紀錄分析了 34 位年輕人的談話。
他們過去曾經是罹患
精神分裂症的高風險族群。
精神分裂症的高風險族群。
09:19
who were at a high risk
of developing schizophrenia.
of developing schizophrenia.
192
547784
2801
09:23
And so what we did is,
we measured speech at day one,
we measured speech at day one,
193
551434
2881
我們測量了他們第一天的談話，
09:26
and then we asked whether the properties
of the speech could predict,
of the speech could predict,
194
554339
3242
然後問電腦，從他們的話中，
是否可以預測出，
是否可以預測出，
09:29
within a window of almost three years,
195
557605
2496
未來三年內，
09:32
the future development of psychosis.
196
560125
2035
他們會不會精神錯亂。
09:35
But despite our hopes,
197
563427
2366
但我們大失所望，
09:37
we got failure after failure.
198
565817
3117
一次又一次的失敗。
09:41
There was just not enough
information in semantics
information in semantics
199
569793
3882
因為沒有足夠的語義資訊
09:45
to predict the future
organization of the mind.
organization of the mind.
200
573699
2793
來預測未來的心智發展。
09:48
It was good enough
201
576516
1809
它在分辨精神病患及控制組上
09:50
to distinguish between a group
of schizophrenics and a control group,
of schizophrenics and a control group,
202
578349
4175
已經有足夠的能力，
09:54
a bit like we had done
for the ancient texts,
for the ancient texts,
203
582548
2712
因為這有點像我們之前
做古文字的分析，
做古文字的分析，
09:57
but not to predict the future
onset of psychosis.
onset of psychosis.
204
585284
2994
但沒辦法預測未來
精神錯亂的發病。
精神錯亂的發病。
10:01
But then we realized
205
589164
1706
後來我們了解到，
10:02
that maybe the most important thing
was not so much what they were saying,
was not so much what they were saying,
206
590894
4088
也許最重要的關鍵
不是他們說了甚麼，
不是他們說了甚麼，
10:07
but how they were saying it.
207
595006
1673
而是他們怎麼說。
10:09
More specifically,
208
597679
1220
更精確地說，
10:10
it was not in which semantic
neighborhoods the words were,
neighborhoods the words were,
209
598923
2827
不是他們說的「話」落在哪個
語義相近的群組裡，
語義相近的群組裡，
10:13
but how far and fast they jumped
210
601774
2600
而是他們說話的「方式」
是否會在這幾個語義相近的群組裡
是否會在這幾個語義相近的群組裡
10:16
from one semantic neighborhood
to the other one.
to the other one.
211
604398
2301
快速地跳來跳去。
10:19
And so we came up with this measure,
212
607247
1731
所以我們想出了一個
10:21
which we termed semantic coherence,
213
609002
2389
叫做「語義連貫性」的量測方法，
10:23
which essentially measures the persistence
of speech within one semantic topic,
of speech within one semantic topic,
214
611415
4804
本質上就是測量談話的持續性
是否會落在同一個
語義主題或類別上。
語義主題或類別上。
10:28
within one semantic category.
215
616243
1529
10:31
And it turned out to be
that for this group of 34 people,
that for this group of 34 people,
216
619294
4047
結果顯示，剛剛的 34 位年輕人，
10:35
the algorithm based on semantic
coherence could predict,
coherence could predict,
217
623365
3659
透過這個語義連貫性演算法，
10:39
with 100 percent accuracy,
218
627048
2500
預測誰會精神錯亂的正確率
達到百分之百。
達到百分之百。
10:41
who developed psychosis and who will not.
219
629572
2507
10:44
And this was something
that could not be achieved --
that could not be achieved --
220
632976
2937
目前臨床上所有測量方式
都無法達到、或接近這個數字。
都無法達到、或接近這個數字。
10:47
not even close --
221
635937
1508
10:49
with all the other
existing clinical measures.
existing clinical measures.
222
637469
3126
10:54
And I remember vividly,
while I was working on this,
while I was working on this,
223
642525
3579
在我做這項研究的時候，
清楚地記得一件事，
清楚地記得一件事，
10:58
I was sitting at my computer
224
646128
2317
當時我坐在電腦前面，
11:00
and I saw a bunch of tweets by Polo --
225
648469
2635
看到之前我回到布宜諾斯艾利斯的第一個學生
——保羅，傳了一堆信息給我，
——保羅，傳了一堆信息給我，
11:03
Polo had been my first student
back in Buenos Aires,
back in Buenos Aires,
226
651128
3167
11:06
and at the time
he was living in New York.
he was living in New York.
227
654319
2070
當時他住在紐約。
11:08
And there was something in this tweets --
228
656413
2088
我發現訊息不太對勁——
11:10
I could not tell exactly what
because nothing was said explicitly --
because nothing was said explicitly --
229
658525
3501
雖然我講不出個所以然來，
因為他寫得不清不楚——
因為他寫得不清不楚——
11:14
but I got this strong hunch,
230
662050
2021
但我有一個強烈的直覺，
一定是出事了。
一定是出事了。
11:16
this strong intuition,
that something was going wrong.
that something was going wrong.
231
664095
2955
11:20
So I picked up the phone,
and I called Polo,
and I called Polo,
232
668347
2723
所以，我打電話給保羅，
11:23
and in fact he was not feeling well.
233
671094
1919
沒錯，他當時感覺不太舒服。
11:25
And this simple fact,
234
673362
1937
用這樣一個單純的辨認方式，
11:27
that reading in between the lines,
235
675323
2491
從他的字裡行間，
11:29
I could sense,
through words, his feelings,
through words, his feelings,
236
677838
4262
我可以隱約感受到他的感覺，
11:34
was a simple, but very
effective way to help.
effective way to help.
237
682124
2619
並在第一時間有效地幫助他。
11:37
What I tell you today
238
685987
1638
今天我要告訴各位的是，
11:39
is that we're getting
close to understanding
close to understanding
239
687649
2508
我們已經越來越能理解
11:42
how we can convert this intuition
that we all have,
that we all have,
240
690181
4286
如何把我們共有的直覺，
轉換成演算法。
轉換成演算法。
11:46
that we all share,
241
694491
1365
11:47
into an algorithm.
242
695880
1197
11:50
And in doing so,
243
698102
1461
經由這樣做，
11:51
we may be seeing in the future
a very different form of mental health,
a very different form of mental health,
244
699587
4650
未來我們也許可以看到一種
全然不同的心理健康模式，
全然不同的心理健康模式，
11:56
based on objective, quantitative
and automated analysis
and automated analysis
245
704261
5621
而且是基於一種客觀、量化的方式
來自動分析出我們所寫的字、
12:01
of the words we write,
246
709906
1709
我們所說的話。
12:03
of the words we say.
247
711639
1537
12:05
Gracias.
248
713200
1151
謝謝。
12:06
(Applause)
249
714375
6883
（掌聲）
ABOUT THE SPEAKER
Mariano Sigman - NeuroscientistIn his provocative, mind-bending book "The Secret Life of the Mind," neuroscientist Mariano Sigman reveals his life’s work exploring the inner workings of the human brain.
Why you should listen
Mariano Sigman, a physicist by training, is a leading figure in the cognitive neuroscience of learning and decision making. Sigman was awarded a Human Frontiers Career Development Award, the National Prize of Physics, the Young Investigator Prize of "College de France," the IBM Scalable Data Analytics Award and is a scholar of the James S. McDonnell Foundation. In 2016 he was made a Laureate of the Pontifical Academy of Sciences.
In The Secret Life of the Mind, Sigman's ambition is to explain the mind so that we can understand ourselves and others more deeply. He shows how we form ideas during our first days of life, how we give shape to our fundamental decisions, how we dream and imagine, why we feel certain emotions, how the brain transforms and how who we are changes with it. Spanning biology, physics, mathematics, psychology, anthropology, linguistics, philosophy and medicine, as well as gastronomy, magic, music, chess, literature and art, The Secret Life of the Mind revolutionizes how neuroscience serves us in our lives, revealing how the infinity of neurons inside our brains manufacture how we perceive, reason, feel, dream and communicate.
Mariano Sigman | Speaker | TED.com