One Down, Many to Go

Administrators at the Georgia Institute of Technology are optimistic but “not declaring victory” after one semester of its affordable online master’s degree program in computer science. While the program has been well-received by students, administrators are still striving to solve an equation that balances cost, academic quality and support services.

“We’re not all the way there yet, but I couldn’t ask for a much better start,” Zvi Galil, dean of the College of Computing, wrote last month in an email to Georgia Tech faculty on the one-year anniversary of the program's announcement.

The initiative has been closely watched since last spring's announcement -- and not just because of the dramatic savings it offers compared to the university’s on-campus program. A three-credit-hour online course costs less than a single credit hour of face-to-face education -- $402 versus $472, based on spring 2013 tuition rates. The goal is to get much larger than a traditional program could sustain, but also much smaller than the average MOOC.

The savings gap may narrow as Georgia Tech scales the program. “We hope to be able to stick with this tuition, but whether this is the right tuition, we don’t know yet,” said Galil, who estimated an enrollment of a few thousand students could be enough to balance the budget.

The master’s degree program also represents an important investment for Udacity, the one-time massive open online course provider that has recently gravitated toward paid certificates and corporate training.

The online computer science program also includes a corporate presence. AT&T has subsidized the program with a $2 million investment (and many students are AT&T employees), which will last the university “until we run out of it -- and we are close!” Galil said. The university is considering similar partnerships with other corporations, he added.

Georgia Tech only admitted 410 students to start this spring and capped the courses at 150 students each, which meant another 350 students who applied last fall had to wait until the summer term, which began on May 19. With another 400 to 600 students likely to be admitted this fall, Galil said the university will continue to scale the program slowly, semester by semester, likely missing an early and optimistic estimate to enroll 10,000 students by the third year.

“We will not accept students just to have large numbers,” Galil said. He later added, “Our philosophy is everybody that is qualified gets admitted. This is counter to the current norm in the elite universities -- the Ivies, the MITs, the Georgia Techs -- where admission is by denial.”

The buzz around the online degree program appears to have benefited the residential program as well. This year, applications were up by 30 percent, the university reported.

More students means more courses. Galil confirmed faculty are developing seven new courses, five of which will be made available this fall. At a rate of four or five new courses per semester, the course catalog will soon offer enough variety to let students to pick specializations such as interactive intelligence, machine learning and social computing, among others. And this summer, the university will offer two not-for-credit tracks.

Some students initially griped about not being able to take more than two courses, but a few weeks into the semester, most of the criticism had died down:

“[Machine learning] is a cool class that will make you want to die. You will wake in cold sweats from nightmares, and you had better have a social support structure as strong as my wife to get through it.”

That’s how one student described his experience on the program’s unofficial community on the social network Reddit. He and 309 out of the 380 students who enrolled this spring “survived” the first semester, as Galil put it. 

In the five courses offered this spring, those students posted a 3.55 grade-point average. When comparing the four courses that were offered both online and on campus, the online students narrowly outperformed their residential counterparts, 3.61 to 3.58. (Sebastian Thrun, the Udacity CEO who teaches the online-only Artificial Intelligence for Robotics, must be a tough grader.)

The student, who goes by the nickname “DrCaret2” on Reddit, spoke to Inside Higher Ed on a condition of anonymity as he is still enrolled in the program. While he was full of praise for the program, saying “It is absolutely, immediately, and unquestionably clear that it is in a league of its own,” he also pointed to some growing pains. Some of the assignments, he said, come with frustratingly vague instructions.

“I think that students -- myself included -- had a hard time adjusting to the vagaries of exploratory learning -- and that’s been an adjustment,” the student said. “A lot of students -- at least in the online program -- we’ve been out of school for longer than residential students have been, and it’s a little different in professional settings, because bosses don’t really do the Socratic lecturing thing that instructors are used to.”

The average student in the online program is 34 years old; the average residential student, 23.

To help one another decipher the assignments, the students have spilled out of the official forum and into not just the Reddit community, but also a Facebook group and a Google+ page. Posts on those sites range from griping about admissions requirements and comparing impressions to discussing Yellow Jackets baseball.

"We are thrilled to see this kind of interaction,” Galil said. “One student berates the fact that it’s difficult, two others jump on him and say, ‘These are Georgia Tech courses.’ This is much better coming from students.”

Each course does come with teaching assistants -- the machine learning course had two, for example -- but they aren’t available around the clock.

“While I appreciate the professor getting back to me, working ahead by 1 week should not be unexpected, and [questions] should be answered,” one student tweeted last month.

Udacity’s role in the program is largely technical -- it provides the platform and course development support, Clarissa Shen, vice president of strategic business and marketing, said in an email. The company was supposed to supply a large number of mentors to assist students, but “certain personnel roles have evolved as the program has taken shape,” she wrote.

Udacity is still involved in finding teaching assistants, suggesting candidates who are then vetted by Georgia Tech. Faculty at the university have also been able to recruit some former students. “It’s not always trivial, and it takes work to find them,” Galil said.

In addition to finding the appropriate number of teaching assistants and graders for each course, Galil said, the university is also pondering how to effectively deliver advising and career counseling. “One of the things we’ve learned is ... to manage expectations,” he said.

Those unresolved matters have yet to put a damper on student enthusiasm, however. In response to a survey, 93 percent of students said they would recommend the program, and 97 percent said it offered a “good or exceptional value” for the cost.

That percentage includes DrCaret2, who said he has been fascinated by MOOCs ever since enrolling in Sebastian Thrun’s first artificial intelligence course in 2011.

“The whole program is a big change from the usual ‘online’ classes offered by other schools,” he wrote. “Another guy at my office is taking a math class at a local school, and comparing the [machine learning] lectures is like watching high school highlight footage of Lebron James before he was drafted.”