In November, Wolfram Burgard, a professor of computer science at the University of Freiburg, in Germany, administered an online midterm exam for a course in artificial intelligence to 54 students. The test-takers sat in the lecture hall, spaced at least a meter apart, with proctors roaming the aisles to make sure nobody was looking up clues or chatting online with co-conspirators.
The students were from all over. Some were enrolled at Freiburg, some at the Technical University of Munich, some at the University of Hamburg, and several from outside Germany. Most were hoping to get credit for the course at their home universities, which meant they would have to return to Freiburg in mid-December to take a proctored final exam; no small chore for a pair visiting from Paris, and the one who had flown in from Finland, a distance of 1,500 miles.
Still, those incurring travel costs could take solace: First, they did not have to trek nearly 6,000 miles to where the course was actually being given, at Stanford University. Second, they stood a chance at getting academic credit for taking the course — previously available only to Stanford students as part of the university’s exclusive, $40,000-per- year-tuition undergraduate curriculum — without paying a dime.
That A.I. course was the flagship of a trio of Stanford computer science courses that were broadcast this fall, for the first time, to anyone on the Internet who cared to log in. This made Stanford the latest of a handful of elite American universities to pull back the curtain on their vaunted courses, joining the Massachusetts Institute of Technology’s OpenCourseWare project, Yale University’s Open Yale Courses and the University of California at Berkeley’s Webcast.Berkeley, among others.
The difference with the Stanford experiment is that students are not only able to view the course materials and tune into recorded lectures for CS221: Introduction to Artificial Intelligence; they are also invited to take in-class quizzes, submit homework assignments, and gather for virtual office hours with the course’s two rock star instructors — Peter Norvig, a research executive at Google who used to build robots for NASA, and Sebastian Thrun, a professor of computer science at Stanford who also works for Google, designing cars that drive themselves. (M.I.T., Yale and Berkeley simply make the course materials freely available, without offering the opportunity to interact with the professors or submit assignments to be graded.)
Including the 54 European students in the University of Freiburg lecture hall, 23,000 far-flung guests took the midterm exam, with many scoring on par with the 175 paying students who took the same test in Palo Alto. Those who also complete the final exam this month will get a letter signed by Thrun, along with their cumulative grade and class rank.
Only the Stanford students, and perhaps some of those taking the tests under Burgard’s watch in Freiburg, stand to redeem their efforts in Norvig and Thrun’s course for credit. And yet the Stanford experiment punches a few more bricks out of the barrier that historically has kept “elite” courses in sequestration, creating more footholds for the so-called “edupunks” who would sooner scale that ivied wall than buy a key to the gate.
“I think it goes a step further [than many existing open courseware projects],” says Thrun. The option of submitting coursework that will be acknowledged by the professors -- rather than just reading a syllabus and watching lectures -- “forces you to exercise,” he says.
“It’s weight loss — not by watching someone else lose weight, but by you doing the work,” says Thrun.
There are other open education projects that promise active learning experiences for non-paying learners, notably Carnegie Mellon University’s Open Learning Initiative (OLI), which allows visitors to interact with a “smart” exercise platform that adapts to their unique needs.
But there are no human lectures or office hours built into the OLI courses. Students taking those courses independently do not get tokens of achievement for completing coursework. The Carnegie Mellon project focuses on foundational courses and is largely oriented to community college curriculums.
Stanford’s project takes aim at more advanced courses. Based on the success of Norvig and Thrun’s experiment, the university’s computer science department is planning to broadcast eight additional courses for free in the spring, most focusing on high-level concepts that require participants already to have a pretty good command of math and science. Norvig and Thrun’s A.I. course, for example, assumes knowledge of linear algebra and probability theory. Next semester’s open courses will include Cryptography, Human Computer Interaction and Probabilistic Graphic Models.
Norvig and Thrun are not the first college professors to experiment with what are known as “massive open online courses,” or MOOCs. Others have been broadcasting their courses to the Web-enabled masses for years.
These include David Wiley, a prominent open education advocate and associate professor of psychology and technology at Brigham Young University. In 2008, as a professor at Utah State University, Wiley opened up a class of 15 to non-enrolled students via the Web, and soon began welcoming contributions from five additional students who were not enrolled at Utah State. Much like Norvig and Thrun, Wiley assessed the work submitted by his online auditors and awarded them unofficial "diplomas" at the conclusion of the course.
Also in 2008, George Siemens, associate director of the Learning Technologies Center at the University of Manitoba, began teaching MOOCs with several thousand participants beyond the “core” of credit-earning Manitoba students. Siemens says he did not always have the means to give the non-enrolled students personalized feedback, but encouraged them to assess each other’s coursework.
“The value points within education are somewhat vague and shifting,” wrote Siemens via e-mail. “M.I.T.’s [OpenCourseWare] questioned the value of content as an economic value point. MOOCs question the value of teaching as an economic value point.”
While the Stanford professors did not invent the MOOC, Norvig and Thrun have lent the idea considerable profile by attaching their names — and by implication, Stanford’s — to the experiment.
Stanford has been careful to make sure its name is left off the tokens of recognition that Norvig and Thrun plan to send to participants who successfully complete the A.I. course. (Utah State officials were likewise adamant that Wiley not include the university’s name anywhere on his do-it-yourself certificates.) Nevertheless, the gesture challenges the idea that meaningful certificates need to bear the seal of a university to have currency with employers.
“I don’t think its significant that ‘Stanford’ is doing this, I think it’s significant that Peter Norvig is doing this,” says Michael Feldstein, a senior program manager for Cengage Learning and author of the popular education technology blog e-Literate. “He’s essentially using his reputation in the field to provide his stamp of approval on a student’s performance, independent of his institution.”
“It raises the question: Whose certification matters, for what purposes?” Feldstein says. “If individual professors can begin to certify student competence, that begins to unravel the entire fabric of the institution itself.” (Feldstein noted that he was speaking independently and not on behalf of Cengage.)
Still, none of the experts I interviewed seriously expects Norvig and Thrun’s experiment to undermine the value proposition of a Stanford degree. There remains unique value in the on-campus experience relative to the marginal opportunities for intimate engagement for the thousands of outsiders peering through the holes Norvig and Thrun have punched in the perimeter of the walled garden.
For one, the professors can only evaluate non-enrolled students via assessments that can be graded automatically. Currently, this means multiple-choice quizzes and exams. In the case of the artificial intelligence course, Thrun says he thinks he can test genuine understanding even within the constraints of multiple-choice, but for other computer science courses — particularly those oriented to programming — it can be difficult to assess skills without being able to administer project-based assignments (although Thrun does say he sees a possible solution in “compilers” — computer programs that evaluate the viability of other programs).
There’s also the fact that Norvig and Thrun, whose good words would certainly count toward anyone’s candidacy for a job in the field, do not plan to personally vouch for any of the non-Stanford contingent. “I’m not going to write 20,000 letters of recommendation,” says Thrun.
“Everyone hyperventilates about [these experiments] replacing traditional education, but I don’t see that happening,” says Steve Carson, director of external relations for M.I.T. OpenCourseWare and former president of the international OpenCourseWare Consortium. “With a player like Stanford doing something like this, they’re bringing attention to the possibilities of the Web for expanding open education,” he says, but “I think it’s going to be a long time before Stanford manages to eat its tail.”
The form letters he and Norvig plan to non-enrolled learners who complete their course might not carry the same weight as a traditional recommendation, Thrun says. "But on the positive side, more artificial intelligence students get a letter from me than all the other professors in the West combined," he says.
For the latest technology news and opinion from Inside Higher Ed, follow @IHEtech on Twitter.