Audio Transcript Auto-generated
- 00:02 - 00:05
this video will present to you to applications for Beijing
- 00:05 - 00:09
networks, starting with intelligent tutoring systems.
- 00:12 - 00:15
Intelligent tutoring systems are computer based tutoring systems.
- 00:16 - 00:18
So we get rid of a human teacher, and instead
- 00:19 - 00:22
we replaced with an artificial intelligence focused on improving your
- 00:23 - 00:24
progress on various tasks.
- 00:25 - 00:27
Why is this something the world needs?
- 00:29 - 00:31
All of you will have experienced the time where you
- 00:32 - 00:35
sit in a room with around 25 other students, all
- 00:35 - 00:37
trying and sometimes struggling toe.
- 00:37 - 00:38
Learn something new.
- 00:39 - 00:40
There is only one teacher.
- 00:40 - 00:44
It is impossible to address every student's individual problems, so
- 00:45 - 00:47
you might be left not understanding a problem and having
- 00:48 - 00:50
to build your knowledge on a foundation with a bunch
- 00:50 - 00:51
of holes.
- 00:52 - 00:55
A study by Bloom has proven that one on one
- 00:55 - 00:58
tutoring is the most effective way for someone to learn.
- 00:59 - 01:01
Also, now that we live in times of a pandemic,
- 01:02 - 01:04
it could come in very handy if there were sufficient
- 01:05 - 01:10
intelligent tutoring systems at hand to educate quarantine students or
- 01:10 - 01:14
allow some students to stay home, uncrowded the classroom while
- 01:14 - 01:16
still allowing sufficient education.
- 01:19 - 01:23
What do intelligent tutoring systems have to provide for individual
- 01:24 - 01:29
tutoring? First of all, the system needs to collect data,
- 01:29 - 01:32
which means while the student is interacting with the system,
- 01:33 - 01:36
his or her behavior needs to be tracked and saved
- 01:37 - 01:38
based on the state.
- 01:38 - 01:40
Um, a model of the student has to be designed
- 01:40 - 01:44
representing the progress and state of understanding he or she
- 01:44 - 01:44
is in.
- 01:46 - 01:48
Based on this model off the student material has to
- 01:49 - 01:51
be selected, which will help him or her to close
- 01:51 - 01:53
the knowledge gaps and solve the tasks.
- 01:55 - 01:58
The student model can never be 100% complete, and the
- 01:58 - 02:00
needed material cannot be determined with certainty.
- 02:01 - 02:03
So we need a concept that can deal with this
- 02:03 - 02:07
type off information and babies.
- 02:07 - 02:11
Networks are capable of doing so because it uses probability
- 02:11 - 02:15
theory to deal with the uncertainty off the student model.
- 02:17 - 02:19
Let's look at a very simplified example.
- 02:22 - 02:25
A student is interacting with the intelligent tutor.
- 02:26 - 02:28
Data is being collected, and the knowledge state off.
- 02:28 - 02:29
The student is being modeled.
- 02:30 - 02:33
Now the student runs into a problem that he or
- 02:34 - 02:35
she cannot solve.
- 02:38 - 02:40
Let's say the problem has two components that you need
- 02:40 - 02:42
to have understood to solve it.
- 02:42 - 02:45
For the data that has been collected, a probability has
- 02:45 - 02:48
been computed which states how likely it is that you
- 02:48 - 02:50
have a problem with component one or two.
- 02:51 - 02:54
The component with the higher probability will be suggested to
- 02:54 - 02:58
you. You rehearse it and again new data will be
- 02:58 - 03:03
collected. Then you return to the problem and I hopefully
- 03:03 - 03:04
capable of solving it.
- 03:05 - 03:08
If not, the system will again suggest the component with
- 03:08 - 03:09
the highest probability to you.
- 03:10 - 03:13
And after rehearsing a certain amount of time, the probability
- 03:13 - 03:16
will be really high, that you can solve the problem.
- 03:18 - 03:21
And this is the basic idea off cohesion networks in
- 03:22 - 03:23
intelligent tutoring systems.