How do 100,000 students take a class together? Does size matter?
This session will show how the CS department at Stanford University is running a course in Artificial Intelligence (AI) for 100,000 students this term.
Please attend if you're curious about the future of such mega-courses and want to know more about the goals, methods and tools of this example and others.
Among the issues discussed will be how participation and feedback are handled in such a large course. Also reported will be the instructor and student experience by the end of the course.
This NY Times article (Aug 15, 2011) provides some background information: Virtual and Artificial, but 58,000 Want Course
I got started with the online AI course when I saw the NY Times piece above.
When I signed up for the course in September the enrollment had jumped from 58,000 to 80,000 students.
Guess how many at its peak?
By the time the course started in October there were over 160,000 students, 5 times the size of UW's student body.
Meet Professors Peter Norvig and Sebastion Thrun from Stanford University:
Inspiration came in part from the
Khan Academy.
(MIT Opencourseware, UK Open University)
From its peak of 160,000 students 190 countries, participation declined to 46,000 students handing in the first homework assignment, and then to 23,000 students by the time of the midterm exam in November. That is 230 times the number of students in UW's or Stanford's AI class on campus.
By Oct 30th the course site reports there were 3,000,000 visits and 11,000,000 page hits.
The video lecture closed-caption text was created in English and a dozen other languages (initially at least) by 2000 (peak) student volunteer translators.
Some course site traffic info from google.
A few comments posted on the course web site ai-class.com show the trend in opinion. Students liked the course. They liked learning from experts and the teaching methods. They asked for more opportunities for similar Stanford courses.
Videos are used for lecture instruction, quizzes, homework, tests and office hours.
Watch here on the course site or on youtube.
The course runs for about 10 weeks and about 20 AI topics. With each topic comprised of 20-30 short videos or 1-7 minutes each.
Quizzes are part of most videos. They are interactive quizzes and are graded for those who have logged into the course. Question types are multiple-choice and fill-in-the-blank.
The quizzes are for practice and mastery not for marks. The 8 homework assignments and the midterm and final exams are also video tests of the same nature. All term work can be repeated anytime. Once the due date is past, students get immediate feedback if they are right or wrong and they can try again if they wish.
A course mark is composed of
Stanford University, while participating in the course, is not granting credit for it. Marks are only for judging performance and understanding.
There is no supervision of the term work. There is nothing a student has to hand in and no work needs to be graded by course staff.
Two other online Stanford courses this term, similar to this one, do have students hand in some course work. See the database and machine learning course sites.
There is an optional textbook, Professor Norvig's Artificial Intelligence: A Modern Approach. Readings from the book are assigned for every course topic.
No email goes to professors. Instead, questions to professors are
submitted to a
course forum.
Students vote for the questions they like or dislike, and
then, professors answer a selection of the top voted questions in
a 20 minute office hour on
youtube.
Students have 2 sites of their own for discussions: aiqus.com and reddit. Here students help each other and share course notes, and code after due dates.
Can you guess? Hint: 10's of thousands of students from around the world.
With so many students in the class, the web servers crashed a couple of times when too many students tried to complete the course term work minutes before the time limits expired. Server capacity was increased each time a problem occurred and due dates were extended.
A successful course. Happy students.
More to come? Definitely. Stanford is rolling out more than a dozen new online courses for the Winter 2012 term. See this collected course list and an example, the Machine Learning course.
In one way or another, this is going to part of the future of education.
Check out more thoughts about Stanford's online courses from computer scientist John Langford at the computer journal CACM in his blog article Somebody's Eating Your Lunch, September 28, 2011.
The course will be finished in a few weeks. I hope to post more here at that time.
Paul Kates
Mathematics Faculty CTE Liaison
pkates@uwaterloo.ca, x37047
Last modification date: Tue Dec 6 19:33:41 2011.