TEDGlobal 2012
Daphne Koller: What we're learning from online education
Filmed:
Readability: 4.8
2,603,043 views
Daphne Koller is enticing top universities to put their most intriguing courses online for free -- not just as a service, but as a way to research how people learn. With Coursera (cofounded by Andrew Ng), each keystroke, quiz, peer-to-peer discussion and self-graded assignment builds an unprecedented pool of data on how knowledge is processed.
Daphne Koller - Educator
With Coursera, Daphne Koller and co-founder Andrew Ng are bringing courses from top colleges online, free, for anyone who wants to take them. Full bio
With Coursera, Daphne Koller and co-founder Andrew Ng are bringing courses from top colleges online, free, for anyone who wants to take them. Full bio
Double-click the English transcript below to play the video.
00:16
Like many of you, I'm one of the lucky people.
0
725
3111
00:19
I was born to a family where education was pervasive.
1
3836
3400
00:23
I'm a third-generation PhD, a daughter of two academics.
2
7236
4238
00:27
In my childhood, I played around in my father's university lab.
3
11474
3794
00:31
So it was taken for granted that I attend some of the best universities,
4
15268
3849
00:35
which in turn opened the door to a world of opportunity.
5
19117
3801
00:38
Unfortunately, most of the people in the world are not so lucky.
6
22918
4120
00:42
In some parts of the world, for example, South Africa,
7
27038
3135
00:46
education is just not readily accessible.
8
30173
2705
00:48
In South Africa, the educational system was constructed
9
32878
2975
00:51
in the days of apartheid for the white minority.
10
35853
2873
00:54
And as a consequence, today there is just not enough spots
11
38726
2700
00:57
for the many more people who want and deserve a high quality education.
12
41426
3852
01:01
That scarcity led to a crisis in January of this year
13
45278
3880
01:05
at the University of Johannesburg.
14
49158
1836
01:06
There were a handful of positions left open
15
50994
2131
01:09
from the standard admissions process, and the night before
16
53125
2969
01:11
they were supposed to open that for registration,
17
56094
2560
01:14
thousands of people lined up outside the gate in a line a mile long,
18
58654
4052
01:18
hoping to be first in line to get one of those positions.
19
62706
3880
01:22
When the gates opened, there was a stampede,
20
66586
2308
01:24
and 20 people were injured and one woman died.
21
68894
3652
01:28
She was a mother who gave her life
22
72546
1940
01:30
trying to get her son a chance at a better life.
23
74486
4063
01:34
But even in parts of the world like the United States
24
78549
3157
01:37
where education is available, it might not be within reach.
25
81706
4356
01:41
There has been much discussed in the last few years
26
86062
2672
01:44
about the rising cost of health care.
27
88734
1989
01:46
What might not be quite as obvious to people
28
90723
2642
01:49
is that during that same period the cost of higher education tuition
29
93365
4022
01:53
has been increasing at almost twice the rate,
30
97387
2480
01:55
for a total of 559 percent since 1985.
31
99867
4280
02:00
This makes education unaffordable for many people.
32
104147
4534
02:04
Finally, even for those who do manage to get the higher education,
33
108681
3801
02:08
the doors of opportunity might not open.
34
112482
2625
02:11
Only a little over half of recent college graduates
35
115107
3207
02:14
in the United States who get a higher education
36
118314
2313
02:16
actually are working in jobs that require that education.
37
120627
3463
02:19
This, of course, is not true for the students
38
124090
1840
02:21
who graduate from the top institutions,
39
125930
1952
02:23
but for many others, they do not get the value
40
127882
2632
02:26
for their time and their effort.
41
130514
3536
02:29
Tom Friedman, in his recent New York Times article,
42
134050
3030
02:32
captured, in the way that no one else could, the spirit behind our effort.
43
137080
4368
02:37
He said the big breakthroughs are what happen
44
141448
3120
02:40
when what is suddenly possible meets what is desperately necessary.
45
144568
3899
02:44
I've talked about what's desperately necessary.
46
148467
2621
02:46
Let's talk about what's suddenly possible.
47
151088
2512
02:49
What's suddenly possible was demonstrated by
48
153600
3119
02:52
three big Stanford classes,
49
156719
1568
02:54
each of which had an enrollment of 100,000 people or more.
50
158287
3880
02:58
So to understand this, let's look at one of those classes,
51
162167
3384
03:01
the Machine Learning class offered by my colleague
52
165551
1920
03:03
and cofounder Andrew Ng.
53
167471
1729
03:05
Andrew teaches one of the bigger Stanford classes.
54
169200
2319
03:07
It's a Machine Learning class,
55
171519
1209
03:08
and it has 400 people enrolled every time it's offered.
56
172728
3518
03:12
When Andrew taught the Machine Learning class to the general public,
57
176246
3265
03:15
it had 100,000 people registered.
58
179511
2616
03:18
So to put that number in perspective,
59
182127
2009
03:20
for Andrew to reach that same size audience
60
184136
2359
03:22
by teaching a Stanford class,
61
186495
1826
03:24
he would have to do that for 250 years.
62
188321
3926
03:28
Of course, he'd get really bored.
63
192247
3486
03:31
So, having seen the impact of this,
64
195733
2737
03:34
Andrew and I decided that we needed to really try and scale this up,
65
198470
3128
03:37
to bring the best quality education to as many people as we could.
66
201598
4120
03:41
So we formed Coursera,
67
205718
1495
03:43
whose goal is to take the best courses
68
207213
3137
03:46
from the best instructors at the best universities
69
210350
3317
03:49
and provide it to everyone around the world for free.
70
213667
4028
03:53
We currently have 43 courses on the platform
71
217695
2600
03:56
from four universities across a range of disciplines,
72
220295
3199
03:59
and let me show you a little bit of an overview
73
223494
1833
04:01
of what that looks like.
74
225327
3278
04:04
(Video) Robert Ghrist: Welcome to Calculus.
75
228605
1213
04:05
Ezekiel Emanuel: Fifty million people are uninsured.
76
229818
1880
04:07
Scott Page: Models help us design more effective institutions and policies.
77
231698
3271
04:10
We get unbelievable segregation.
78
234969
2408
04:13
Scott Klemmer: So Bush imagined that in the future,
79
237377
1792
04:15
you'd wear a camera right in the center of your head.
80
239169
2378
04:17
Mitchell Duneier: Mills wants the student of sociology to develop the quality of mind ...
81
241547
4254
04:21
RG: Hanging cable takes on the form of a hyperbolic cosine.
82
245801
3665
04:25
Nick Parlante: For each pixel in the image, set the red to zero.
83
249466
3071
04:28
Paul Offit: ... Vaccine allowed us to eliminate polio virus.
84
252537
2977
04:31
Dan Jurafsky: Does Lufthansa serve breakfast and San Jose? Well, that sounds funny.
85
255514
3623
04:35
Daphne Koller: So this is which coin you pick, and this is the two tosses.
86
259137
3616
04:38
Andrew Ng: So in large-scale machine learning, we'd like to come up with computational ...
87
262753
3687
04:42
(Applause)
88
266440
5609
04:47
DK: It turns out, maybe not surprisingly,
89
272049
2274
04:50
that students like getting the best content
90
274323
2238
04:52
from the best universities for free.
91
276561
2887
04:55
Since we opened the website in February,
92
279448
2522
04:57
we now have 640,000 students from 190 countries.
93
281970
4358
05:02
We have 1.5 million enrollments,
94
286328
2152
05:04
6 million quizzes in the 15 classes that have launched
95
288480
2850
05:07
so far have been submitted, and 14 million videos have been viewed.
96
291330
4916
05:12
But it's not just about the numbers,
97
296246
2518
05:14
it's also about the people.
98
298764
1641
05:16
Whether it's Akash, who comes from a small town in India
99
300405
2976
05:19
and would never have access in this case
100
303381
2175
05:21
to a Stanford-quality course
101
305556
1489
05:22
and would never be able to afford it.
102
307045
2515
05:25
Or Jenny, who is a single mother of two
103
309560
2038
05:27
and wants to hone her skills
104
311598
1967
05:29
so that she can go back and complete her master's degree.
105
313565
3135
05:32
Or Ryan, who can't go to school,
106
316700
3136
05:35
because his immune deficient daughter
107
319836
1865
05:37
can't be risked to have germs come into the house,
108
321701
3383
05:40
so he couldn't leave the house.
109
325084
1840
05:42
I'm really glad to say --
110
326924
1632
05:44
recently, we've been in correspondence with Ryan --
111
328556
2252
05:46
that this story had a happy ending.
112
330808
1932
05:48
Baby Shannon -- you can see her on the left --
113
332740
1903
05:50
is doing much better now,
114
334643
1351
05:51
and Ryan got a job by taking some of our courses.
115
335994
4198
05:56
So what made these courses so different?
116
340192
2244
05:58
After all, online course content has been available for a while.
117
342436
3720
06:02
What made it different was that this was real course experience.
118
346156
3712
06:05
It started on a given day,
119
349868
1726
06:07
and then the students would watch videos on a weekly basis
120
351594
3634
06:11
and do homework assignments.
121
355228
1855
06:12
And these would be real homework assignments
122
357083
1791
06:14
for a real grade, with a real deadline.
123
358874
3304
06:18
You can see the deadlines and the usage graph.
124
362178
2056
06:20
These are the spikes showing
125
364234
2088
06:22
that procrastination is global phenomenon.
126
366322
3789
06:26
(Laughter)
127
370111
2576
06:28
At the end of the course,
128
372687
1672
06:30
the students got a certificate.
129
374359
1856
06:32
They could present that certificate
130
376215
2160
06:34
to a prospective employer and get a better job,
131
378375
2153
06:36
and we know many students who did.
132
380528
2060
06:38
Some students took their certificate
133
382588
1919
06:40
and presented this to an educational institution at which they were enrolled
134
384507
3122
06:43
for actual college credit.
135
387629
1841
06:45
So these students were really getting something meaningful
136
389470
2214
06:47
for their investment of time and effort.
137
391684
2834
06:50
Let's talk a little bit about some of the components
138
394518
2555
06:52
that go into these courses.
139
397073
1892
06:54
The first component is that when you move away
140
398965
2628
06:57
from the constraints of a physical classroom
141
401593
2297
06:59
and design content explicitly for an online format,
142
403890
2840
07:02
you can break away from, for example,
143
406730
2528
07:05
the monolithic one-hour lecture.
144
409258
2415
07:07
You can break up the material, for example,
145
411673
1785
07:09
into these short, modular units of eight to 12 minutes,
146
413458
3376
07:12
each of which represents a coherent concept.
147
416834
2974
07:15
Students can traverse this material in different ways,
148
419808
2570
07:18
depending on their background, their skills or their interests.
149
422378
3704
07:21
So, for example, some students might benefit
150
426082
2520
07:24
from a little bit of preparatory material
151
428602
2760
07:27
that other students might already have.
152
431362
2071
07:29
Other students might be interested in a particular
153
433433
2440
07:31
enrichment topic that they want to pursue individually.
154
435873
3086
07:34
So this format allows us to break away
155
438959
3235
07:38
from the one-size-fits-all model of education,
156
442194
2824
07:40
and allows students to follow a much more personalized curriculum.
157
445018
3992
07:44
Of course, we all know as educators
158
449010
2343
07:47
that students don't learn by sitting and passively watching videos.
159
451353
3360
07:50
Perhaps one of the biggest components of this effort
160
454713
2945
07:53
is that we need to have students
161
457658
2592
07:56
who practice with the material
162
460250
2409
07:58
in order to really understand it.
163
462659
3156
08:01
There's been a range of studies that demonstrate the importance of this.
164
465815
3268
08:04
This one that appeared in Science last year, for example,
165
469083
2532
08:07
demonstrates that even simple retrieval practice,
166
471615
2832
08:10
where students are just supposed to repeat
167
474447
2792
08:13
what they already learned
168
477239
1400
08:14
gives considerably improved results
169
478639
1920
08:16
on various achievement tests down the line
170
480559
2269
08:18
than many other educational interventions.
171
482828
4304
08:23
We've tried to build in retrieval practice into the platform,
172
487132
2962
08:25
as well as other forms of practice in many ways.
173
490094
2254
08:28
For example, even our videos are not just videos.
174
492348
4144
08:32
Every few minutes, the video pauses
175
496492
2043
08:34
and the students get asked a question.
176
498535
2151
08:36
(Video) SP: ... These four things. Prospect theory, hyperbolic discounting,
177
500686
2221
08:38
status quo bias, base rate bias. They're all well documented.
178
502907
3092
08:41
So they're all well documented deviations from rational behavior.
179
505999
2767
08:44
DK: So here the video pauses,
180
508766
1624
08:46
and the student types in the answer into the box
181
510390
2256
08:48
and submits. Obviously they weren't paying attention.
182
512646
3223
08:51
(Laughter)
183
515884
869
08:52
So they get to try again,
184
516753
2010
08:54
and this time they got it right.
185
518763
2536
08:57
There's an optional explanation if they want.
186
521299
2193
08:59
And now the video moves on to the next part of the lecture.
187
523492
4257
09:03
This is a kind of simple question
188
527749
1878
09:05
that I as an instructor might ask in class,
189
529627
2081
09:07
but when I ask that kind of a question in class,
190
531708
2500
09:10
80 percent of the students
191
534208
1300
09:11
are still scribbling the last thing I said,
192
535508
1866
09:13
15 percent are zoned out on Facebook,
193
537374
3321
09:16
and then there's the smarty pants in the front row
194
540695
2456
09:19
who blurts out the answer
195
543151
1359
09:20
before anyone else has had a chance to think about it,
196
544510
2207
09:22
and I as the instructor am terribly gratified
197
546717
2872
09:25
that somebody actually knew the answer.
198
549589
1648
09:27
And so the lecture moves on before, really,
199
551237
2792
09:29
most of the students have even noticed that a question had been asked.
200
554029
3529
09:33
Here, every single student
201
557558
2607
09:36
has to engage with the material.
202
560165
2784
09:38
And of course these simple retrieval questions
203
562949
1936
09:40
are not the end of the story.
204
564885
1662
09:42
One needs to build in much more meaningful practice questions,
205
566547
2970
09:45
and one also needs to provide the students with feedback
206
569517
2353
09:47
on those questions.
207
571870
1663
09:49
Now, how do you grade the work of 100,000 students
208
573533
2888
09:52
if you do not have 10,000 TAs?
209
576421
3082
09:55
The answer is, you need to use technology
210
579503
2354
09:57
to do it for you.
211
581857
1495
09:59
Now, fortunately, technology has come a long way,
212
583352
2648
10:01
and we can now grade a range of interesting types of homework.
213
586000
3268
10:05
In addition to multiple choice
214
589268
1527
10:06
and the kinds of short answer questions that you saw in the video,
215
590795
3153
10:09
we can also grade math, mathematical expressions
216
593948
3260
10:13
as well as mathematical derivations.
217
597208
1952
10:15
We can grade models, whether it's
218
599160
2874
10:17
financial models in a business class
219
602034
2176
10:20
or physical models in a science or engineering class
220
604210
2984
10:23
and we can grade some pretty sophisticated programming assignments.
221
607194
3744
10:26
Let me show you one that's actually pretty simple
222
610938
1919
10:28
but fairly visual.
223
612857
1480
10:30
This is from Stanford's Computer Science 101 class,
224
614337
2477
10:32
and the students are supposed to color-correct
225
616814
1604
10:34
that blurry red image.
226
618418
1592
10:35
They're typing their program into the browser,
227
620010
2018
10:37
and you can see they didn't get it quite right, Lady Liberty is still seasick.
228
622028
4058
10:41
And so, the student tries again, and now they got it right, and they're told that,
229
626086
3756
10:45
and they can move on to the next assignment.
230
629842
2359
10:48
This ability to interact actively with the material
231
632201
3148
10:51
and be told when you're right or wrong
232
635349
1684
10:52
is really essential to student learning.
233
637033
3126
10:56
Now, of course we cannot yet grade
234
640159
2275
10:58
the range of work that one needs for all courses.
235
642434
2834
11:01
Specifically, what's lacking is the kind of critical thinking work
236
645268
3301
11:04
that is so essential in such disciplines
237
648569
1922
11:06
as the humanities, the social sciences, business and others.
238
650491
3597
11:09
So we tried to convince, for example,
239
654088
2249
11:12
some of our humanities faculty
240
656337
1616
11:13
that multiple choice was not such a bad strategy.
241
657953
2696
11:16
That didn't go over really well.
242
660649
2191
11:18
So we had to come up with a different solution.
243
662840
2433
11:21
And the solution we ended up using is peer grading.
244
665273
3074
11:24
It turns out that previous studies show,
245
668347
2422
11:26
like this one by Saddler and Good,
246
670769
1672
11:28
that peer grading is a surprisingly effective strategy
247
672441
2488
11:30
for providing reproducible grades.
248
674929
3214
11:34
It was tried only in small classes,
249
678143
1770
11:35
but there it showed, for example,
250
679913
1487
11:37
that these student-assigned grades on the y-axis
251
681400
2482
11:39
are actually very well correlated
252
683882
1311
11:41
with the teacher-assigned grade on the x-axis.
253
685193
2296
11:43
What's even more surprising is that self-grades,
254
687489
3160
11:46
where the students grade their own work critically --
255
690649
2311
11:48
so long as you incentivize them properly
256
692960
1737
11:50
so they can't give themselves a perfect score --
257
694697
1938
11:52
are actually even better correlated with the teacher grades.
258
696635
3191
11:55
And so this is an effective strategy
259
699826
1607
11:57
that can be used for grading at scale,
260
701433
2104
11:59
and is also a useful learning strategy for the students,
261
703537
2736
12:02
because they actually learn from the experience.
262
706273
2255
12:04
So we now have the largest peer-grading pipeline ever devised,
263
708528
4649
12:09
where tens of thousands of students
264
713177
2504
12:11
are grading each other's work,
265
715681
1198
12:12
and quite successfully, I have to say.
266
716879
3069
12:15
But this is not just about students
267
719948
2260
12:18
sitting alone in their living room working through problems.
268
722208
3041
12:21
Around each one of our courses,
269
725249
1807
12:22
a community of students had formed,
270
727056
2160
12:25
a global community of people
271
729216
1880
12:26
around a shared intellectual endeavor.
272
731096
2532
12:29
What you see here is a self-generated map
273
733628
2652
12:32
from students in our Princeton Sociology 101 course,
274
736280
2961
12:35
where they have put themselves on a world map,
275
739241
2759
12:37
and you can really see the global reach of this kind of effort.
276
742000
2960
12:40
Students collaborated in these courses in a variety of different ways.
277
744960
4567
12:45
First of all, there was a question and answer forum,
278
749527
2639
12:48
where students would pose questions,
279
752166
2144
12:50
and other students would answer those questions.
280
754310
2424
12:52
And the really amazing thing is,
281
756734
1713
12:54
because there were so many students,
282
758447
1670
12:56
it means that even if a student posed a question
283
760117
2365
12:58
at 3 o'clock in the morning,
284
762482
1632
13:00
somewhere around the world,
285
764114
1582
13:01
there would be somebody who was awake
286
765696
2074
13:03
and working on the same problem.
287
767770
2313
13:05
And so, in many of our courses,
288
770083
1958
13:07
the median response time for a question
289
772041
2329
13:10
on the question and answer forum was 22 minutes.
290
774370
3418
13:13
Which is not a level of service I have ever offered to my Stanford students.
291
777788
4577
13:18
(Laughter)
292
782365
1341
13:19
And you can see from the student testimonials
293
783706
1942
13:21
that students actually find
294
785648
1687
13:23
that because of this large online community,
295
787335
2521
13:25
they got to interact with each other in many ways
296
789856
2599
13:28
that were deeper than they did in the context of the physical classroom.
297
792455
4193
13:32
Students also self-assembled,
298
796648
2344
13:34
without any kind of intervention from us,
299
798992
1863
13:36
into small study groups.
300
800855
1903
13:38
Some of these were physical study groups
301
802758
2362
13:41
along geographical constraints
302
805120
1826
13:42
and met on a weekly basis to work through problem sets.
303
806946
2722
13:45
This is the San Francisco study group,
304
809668
1900
13:47
but there were ones all over the world.
305
811568
2319
13:49
Others were virtual study groups,
306
813887
2032
13:51
sometimes along language lines or along cultural lines,
307
815919
2989
13:54
and on the bottom left there,
308
818908
1444
13:56
you see our multicultural universal study group
309
820352
3796
14:00
where people explicitly wanted to connect
310
824148
1763
14:01
with people from other cultures.
311
825911
3006
14:04
There are some tremendous opportunities
312
828917
2111
14:06
to be had from this kind of framework.
313
831028
3325
14:10
The first is that it has the potential of giving us
314
834353
3654
14:13
a completely unprecedented look
315
838007
2434
14:16
into understanding human learning.
316
840441
2289
14:18
Because the data that we can collect here is unique.
317
842730
3463
14:22
You can collect every click, every homework submission,
318
846193
4009
14:26
every forum post from tens of thousands of students.
319
850202
4363
14:30
So you can turn the study of human learning
320
854565
2343
14:32
from the hypothesis-driven mode
321
856908
1933
14:34
to the data-driven mode, a transformation that,
322
858841
2858
14:37
for example, has revolutionized biology.
323
861699
3041
14:40
You can use these data to understand fundamental questions
324
864740
3424
14:44
like, what are good learning strategies
325
868164
1880
14:45
that are effective versus ones that are not?
326
870044
2696
14:48
And in the context of particular courses,
327
872740
2240
14:50
you can ask questions
328
874980
1537
14:52
like, what are some of the misconceptions that are more common
329
876517
3255
14:55
and how do we help students fix them?
330
879772
2177
14:57
So here's an example of that,
331
881949
1424
14:59
also from Andrew's Machine Learning class.
332
883373
2016
15:01
This is a distribution of wrong answers
333
885389
2208
15:03
to one of Andrew's assignments.
334
887597
1610
15:05
The answers happen to be pairs of numbers,
335
889207
1893
15:07
so you can draw them on this two-dimensional plot.
336
891100
2271
15:09
Each of the little crosses that you see is a different wrong answer.
337
893371
3778
15:13
The big cross at the top left
338
897149
2406
15:15
is where 2,000 students
339
899555
2148
15:17
gave the exact same wrong answer.
340
901703
3045
15:20
Now, if two students in a class of 100
341
904748
2327
15:22
give the same wrong answer,
342
907075
1287
15:24
you would never notice.
343
908362
1351
15:25
But when 2,000 students give the same wrong answer,
344
909713
2560
15:28
it's kind of hard to miss.
345
912273
1697
15:29
So Andrew and his students went in,
346
913970
2192
15:32
looked at some of those assignments,
347
916162
1520
15:33
understood the root cause of the misconception,
348
917682
4088
15:37
and then they produced a targeted error message
349
921770
2520
15:40
that would be provided to every student
350
924290
2249
15:42
whose answer fell into that bucket,
351
926539
2179
15:44
which means that students who made that same mistake
352
928718
2084
15:46
would now get personalized feedback
353
930802
2026
15:48
telling them how to fix their misconception much more effectively.
354
932828
4399
15:53
So this personalization is something that one can then build
355
937227
3811
15:56
by having the virtue of large numbers.
356
941038
3140
16:00
Personalization is perhaps
357
944178
2312
16:02
one of the biggest opportunities here as well,
358
946490
2423
16:04
because it provides us with the potential
359
948913
2345
16:07
of solving a 30-year-old problem.
360
951258
2690
16:09
Educational researcher Benjamin Bloom, in 1984,
361
953948
3349
16:13
posed what's called the 2 sigma problem,
362
957297
2251
16:15
which he observed by studying three populations.
363
959548
3062
16:18
The first is the population that studied in a lecture-based classroom.
364
962610
3608
16:22
The second is a population of students that studied
365
966218
2777
16:24
using a standard lecture-based classroom,
366
968995
1719
16:26
but with a mastery-based approach,
367
970714
2080
16:28
so the students couldn't move on to the next topic
368
972794
1920
16:30
before demonstrating mastery of the previous one.
369
974714
3354
16:33
And finally, there was a population of students
370
978068
2294
16:36
that were taught in a one-on-one instruction using a tutor.
371
980362
4528
16:40
The mastery-based population was a full standard deviation,
372
984890
3272
16:44
or sigma, in achievement scores better
373
988162
2288
16:46
than the standard lecture-based class,
374
990450
2394
16:48
and the individual tutoring gives you 2 sigma
375
992844
2144
16:50
improvement in performance.
376
994988
1830
16:52
To understand what that means,
377
996818
1463
16:54
let's look at the lecture-based classroom,
378
998281
1833
16:56
and let's pick the median performance as a threshold.
379
1000114
2919
16:58
So in a lecture-based class,
380
1003033
1338
17:00
half the students are above that level and half are below.
381
1004371
3879
17:04
In the individual tutoring instruction,
382
1008250
2098
17:06
98 percent of the students are going to be above that threshold.
383
1010348
4801
17:11
Imagine if we could teach so that 98 percent of our students
384
1015149
3920
17:14
would be above average.
385
1019069
2198
17:17
Hence, the 2 sigma problem.
386
1021267
3423
17:20
Because we cannot afford, as a society,
387
1024690
2399
17:22
to provide every student with an individual human tutor.
388
1027089
3072
17:26
But maybe we can afford to provide each student
389
1030161
2249
17:28
with a computer or a smartphone.
390
1032410
2019
17:30
So the question is, how can we use technology
391
1034429
2189
17:32
to push from the left side of the graph, from the blue curve,
392
1036618
3375
17:35
to the right side with the green curve?
393
1039993
2738
17:38
Mastery is easy to achieve using a computer,
394
1042731
2337
17:40
because a computer doesn't get tired
395
1045068
1405
17:42
of showing you the same video five times.
396
1046473
3073
17:45
And it doesn't even get tired of grading the same work multiple times,
397
1049546
3251
17:48
we've seen that in many of the examples that I've shown you.
398
1052802
3026
17:51
And even personalization
399
1055828
1854
17:53
is something that we're starting to see the beginnings of,
400
1057682
2136
17:55
whether it's via the personalized trajectory through the curriculum
401
1059818
3192
17:58
or some of the personalized feedback that we've shown you.
402
1063010
3264
18:02
So the goal here is to try and push,
403
1066274
2488
18:04
and see how far we can get towards the green curve.
404
1068762
3497
18:08
So, if this is so great, are universities now obsolete?
405
1072259
5359
18:13
Well, Mark Twain certainly thought so.
406
1077618
2992
18:16
He said that, "College is a place where a professor's lecture notes
407
1080610
2545
18:19
go straight to the students' lecture notes,
408
1083155
1703
18:20
without passing through the brains of either."
409
1084858
2376
18:23
(Laughter)
410
1087234
4047
18:27
I beg to differ with Mark Twain, though.
411
1091281
2668
18:29
I think what he was complaining about is not
412
1093949
2665
18:32
universities but rather the lecture-based format
413
1096614
2750
18:35
that so many universities spend so much time on.
414
1099364
2784
18:38
So let's go back even further, to Plutarch,
415
1102148
3159
18:41
who said that, "The mind is not a vessel that needs filling,
416
1105307
2227
18:43
but wood that needs igniting."
417
1107534
2023
18:45
And maybe we should spend less time at universities
418
1109557
2190
18:47
filling our students' minds with content
419
1111747
2571
18:50
by lecturing at them, and more time igniting their creativity,
420
1114318
3800
18:54
their imagination and their problem-solving skills
421
1118118
3255
18:57
by actually talking with them.
422
1121373
2498
18:59
So how do we do that?
423
1123871
1367
19:01
We do that by doing active learning in the classroom.
424
1125238
3431
19:04
So there's been many studies, including this one,
425
1128669
2449
19:07
that show that if you use active learning,
426
1131118
2080
19:09
interacting with your students in the classroom,
427
1133198
2416
19:11
performance improves on every single metric --
428
1135614
2696
19:14
on attendance, on engagement and on learning
429
1138310
2449
19:16
as measured by a standardized test.
430
1140759
2055
19:18
You can see, for example, that the achievement score
431
1142814
1864
19:20
almost doubles in this particular experiment.
432
1144678
2870
19:23
So maybe this is how we should spend our time at universities.
433
1147548
4401
19:27
So to summarize, if we could offer a top quality education
434
1151949
4577
19:32
to everyone around the world for free,
435
1156526
1903
19:34
what would that do? Three things.
436
1158429
2821
19:37
First it would establish education as a fundamental human right,
437
1161250
3421
19:40
where anyone around the world
438
1164671
1366
19:41
with the ability and the motivation
439
1166037
1921
19:43
could get the skills that they need
440
1167958
1951
19:45
to make a better life for themselves,
441
1169909
1585
19:47
their families and their communities.
442
1171494
2017
19:49
Second, it would enable lifelong learning.
443
1173511
2631
19:52
It's a shame that for so many people,
444
1176142
1951
19:53
learning stops when we finish high school or when we finish college.
445
1178093
3312
19:57
By having this amazing content be available,
446
1181405
2481
19:59
we would be able to learn something new
447
1183886
2743
20:02
every time we wanted,
448
1186629
1136
20:03
whether it's just to expand our minds
449
1187765
1329
20:04
or it's to change our lives.
450
1189094
1959
20:06
And finally, this would enable a wave of innovation,
451
1191053
3145
20:10
because amazing talent can be found anywhere.
452
1194198
3072
20:13
Maybe the next Albert Einstein or the next Steve Jobs
453
1197270
3008
20:16
is living somewhere in a remote village in Africa.
454
1200278
2615
20:18
And if we could offer that person an education,
455
1202893
2656
20:21
they would be able to come up with the next big idea
456
1205549
2356
20:23
and make the world a better place for all of us.
457
1207905
2404
20:26
Thank you very much.
458
1210309
1160
20:27
(Applause)
459
1211469
7583
ABOUT THE SPEAKER
Daphne Koller - EducatorWith Coursera, Daphne Koller and co-founder Andrew Ng are bringing courses from top colleges online, free, for anyone who wants to take them.
Why you should listen
A 3rd generation Ph.D who is passionate about education, Stanford professor Daphne Koller is excited to be making the college experience available to anyone through her startup, Coursera. With classes from 85 top colleges, Coursera is an innovative model for online learning. While top schools have been putting lectures online for years, Coursera's platform supports the other vital aspect of the classroom: tests and assignments that reinforce learning.
At the Stanford Artificial Intelligence Laboratory, computer scientist Daphne Koller studies how to model large, complicated decisions with lots of uncertainty. (Her research group is called DAGS, which stands for Daphne's Approximate Group of Students.) In 2004, she won a MacArthur Fellowship for her work, which involves, among other things, using Bayesian networks and other techniques to explore biomedical and genetic data sets.
Daphne Koller | Speaker | TED.com