CAPPS - Avocacy and Communication Professional Development

California Association of Private Postsecondary Schools

Tempering the Rise of the Machines

05/02/2012

Inside Higher Ed, May 1, 2012

The machines are rising. Soon they will be sophisticated enough to fill certain faculty roles at traditional universities. But to make this revolution work for students, academic leaders at those traditional institutions will need to broker a peace between artificially intelligent teaching programs and their human counterparts, according to a new report written by the former presidents of two prominent traditional universities on behalf of the nonprofit Ithaka S+R.

Online education has enabled many colleges to transition into the prevailing modern medium while adding new sources of revenue in times of scarcity, according to the Ithaka report. However, these innovative colleges have shown less interest in using the novel medium to curb tuition charges and measure learning outcomes.

The report, called "Barriers to Adoption of Online Learning Systems in U.S. Higher Education," was co-written by Lawrence S. Bacow and William G. Bowen, the former presidents of Tufts and Princeton Universities, respectively, along with several Ithaka analysts. It was bankrolled by the Bill & Melinda Gates Foundation. The report contained little advocacy one way or another; rather, the authors appeared to strive for a dispassionate analysis driven by a general sense that the rise of machine learning is inevitable and universities should be prepared. Their findings were based on interviews with senior administrators at 25 public and private, four-year and two-year colleges, including “deep dive” analyses at five of them.

Their objective was to assess the potential roadblocks that might prevent these traditional institutions from adopting sophisticated, “machine guided” learning tools into their curriculums. Technology designed to usher students through new material is thought likely to play a significant role in the future of higher education, although critics have worried that relying too heavily on such technology could harm learning.

In 2009, a team of machine-learning researchers for Carnegie Mellon University’s Online Learning Initiative tested autonomous software that taught a statistics course twice as efficiently as a human lecturer. Companies such as Pearson and Knewton have developed tutoring software that senses weaknesses in the students’ understanding and adapts to their needs. Engineers at Khan Academy, the free learning website, have built similarly adaptive tools into the companion exercises for Khan’s popular video tutorials. The Massachusetts Institute of Technology says it wants to develop virtual electrical engineering laboratories where students -- hundreds of thousands at a time -- can experiment with circuits.

The hope is that these systems will not only help teach students new concepts but also will aid their human instructors by collecting data on how students interact with those ideas and help identify, on a student-by-student basis, what sort of human intervention might be helpful. Rather than replacing professors, ILO technology could equip instructors with the information to do their jobs better. Or so the thinking goes.

Bacow, Bowen and their Ithaka co-authors coin a term in their report to describe this technology: “Interactive Learning Online” systems, or ILO. “Full implementation of sophisticated ILO systems where instruction is either exclusively or largely machine-guided remains quite rare,” they observe. But it will soon shake the windows and rattle the walls of traditional universities, and it will be up to administrators to come up with an adoption strategy that does not shortchange students or start a war with faculty members, the former presidents write.

The authors predict that “a wide variety of such systems, of varying quality and sophistication, will proliferate in the next three to five years,” changing the face of traditional higher education as they do. But based on how colleges and universities such as those that they interviewed have adopted other forms of online education, it is not clear whether the rise of the machines will correlate with a rise in institutional accountability or a fall in tuition prices.

“We believe that [ILO] technology will bring about fundamental reform in how teachers teach and learners learn in years to come,” they write. “Whether these reforms also significantly lower the cost of education remains an open question."

The authors arrive at this conclusion by examining what happened when traditional institutions began integrating the current regime of online education technology. The most frequently cited motivation for adopting online programs was “the desire to generate new revenue streams,” they write. Public universities have used that additional revenue to “offset declines in public appropriations or to supplement faculty compensation.” Private colleges have used it to “address budgetary shortfalls or, in some cases, to directly support traditional modes of instruction.”

However, “Very few are using either savings from online education or the net incremental revenue to reduce the price of education to students,” the authors state.

Neither has a belief in superior pedagogical opportunities played much of a role in the decision of traditional institutions to push their curriculums on to the Web.

“Aside from a few institutions’ references to improvements in retention or pass rates, most interviewees did not explicitly mention a desire for better learning outcomes as a main factor behind their decisions to increase their online offerings,” write Bacow and Bowen. To the contrary, “the belief that students in online courses may learn the material better than their traditional-format counterparts did not appear to be widely held.”

The data that could be harnessed by ILO systems as they proliferate over the next half-decade could lead to a greater understanding of learning outcomes and perhaps greater institutional accountability. But there are potential obstacles to the widespread adoption of such systems -- most revolving around faculty skepticism -- that the authors enumerate.

Prominent among them is the perennial fear that instructors could lose their jobs to interloping automatons. Beyond that, intellectual property issues might be the thorniest obstacle, the authors argue. “Faculty are extremely reluctant to teach courses that they do not ‘own,’ ” the authors write. And “The familiar textbook model in which faculty authors retain copyright does not always translate well for online courses developed with institutional support that may run into the hundreds of thousands of dollars.”

As for using prefab machine-guided courses developed by outsiders, such as Pearson or Carnegie Mellon, professors might not be keen to use automated tutorials that they cannot add to and remix based on their own styles and tastes. Among Ithaka’s interviewees, the authors write, “There was a uniform assertion at all types of institutions that faculty feel much better about teaching repurposed courses or reusing course materials created elsewhere if they are able to do some customization.”

Providing a way for instructors to “brand courses as their own” is the most glaring barrier to machine-learning adoption at traditional universities, according to the report. Inconveniently, it might also be the most difficult to solve.

“To date, no sustainable platform exists that allows interested faculty either to create a fully interactive, machine-guided learning environment or to customize a course that has been created by someone else (and thus claim it as their own),” Bacow and Bowen write. “This is perhaps the largest obstacle to widespread adoption of ILO-style courses.”