Over the past few years, news of the success of data-fed virtual teaching assistants and smart enrollment counselor chatbots has had the higher education world abuzz with the possibilities inherent in using artificial intelligence on campus.
Colleges and universities hope AI will help them offload time-intensive administrative and academic tasks, make IT processes more efficient, boost enrollment in a climate of decline and deliver a better learning experience for students. On some campuses, these improvements are already taking place.
While scaling up AI deployment at universities will take time due to the costs involved, some faculty members may also be resistant to AI on campus because they worry it will put them out of their jobs.
The best way to convince potential stakeholders of the need for AI is to “opt for a problem-first approach,” suggests the Education Advisory Board, an education enrollment services and research company.
“Market machine learning as a solution to strategic imperatives rather than just another flashy technology gimmick,” the company adds.
Another way to convince stakeholders is to highlight stories of successful AI deployments that also demonstrate tangible benefits. Let’s look at a few of those.
How Georgia Tech Used AI to Unburden Harried Teaching Assistants
At the Georgia Institute of Technology, many of the students in a master’s-level AI class were unaware that one of their teaching assistants, Jill Watson, wasn’t human. (This was despite the clue in her name, which refers to IBM’s Watson.)
The class’s approximately 300 students posted about 10,000 messages a semester to an online message board, a volume nearly impossible for a regular assistant to handle, according to The Wall Street Journal. The class’s professor, Ashok Goel, tells Slate that while “the number of questions increases if you have more students … the number of different questions doesn’t really go up." So, he and his team created a system that could respond to those queries that were consistently repeated, and released Jill onto the message board. They populated Jill’s memory with tens of thousands of questions (and their answers) from past semesters.
Not only did most students not realize Jill was virtual, she was also among the most effective teaching assistants the class had seen, answering questions with a 97 percent success rate, according to Slate. Jill had “learned to parse the context of queries and reply to them accurately.”
Jill was one of nine teaching assistants for the course, and her success didn’t mean all the assistants would those their jobs. She couldn’t answer all of the questions — but more important, she couldn’t motivate students or help them with coursework. What Jill did was free up the human teaching assistants to do more meaningful work.
“Where humans cannot go, Jill will go. And what humans do not want to do, Jill can automate,” Goel tells DOGO News.
AI Freezes Summer Melt at Georgia State
In 2016, Georgia State University introduced an AI chatbot, Pounce, that reduced “summer melt” by 22 percent, which meant 324 additional students showed up for the first day of fall classes. “Summer melt” occurs when students who enroll in the spring drop out by the time school begins in the fall. Georgia State’s freshman gains came specifically from those students who had access to the chatbot in a randomized control trial, said the university in a statement.
How did this happen? Through smart text messaging.
The university already knew the advantages of communicating with students via text messages. It also was aware that its existing staff couldn’t possibly be burdened with texting answers to thousands of student queries, according to Campus Technology. It decided to partner with Boston-based AdmitHub, an education technology company that works on conversational AI technology powered by human expertise.
Almost 60 percent of Georgia State’s students are from low-income backgrounds, and many of them are the first in their families to attend college, so they need individual attention as well as financial aid, according to The Chronicle of Higher Education. Feeling confused by various required forms and not knowing which campus offices to go to for specific queries are among the reasons they don’t make it to the first day of classes in the fall.
AdmitHub worked with Georgia State’s admissions and IT teams to identify these and other common obstacles to enrollment that students face, including financial aid applications, placement exams and class registration. Information and answers related to all of these subjects were fed into Pounce, and students could ask Pounce questions 24/7 via text messages on their smart devices.
In 2016, during the first summer of implementation in the randomized control trial, Pounce delivered more than 200,000 answers to questions asked by incoming freshmen.
“Every interaction was tailored to the specific student’s enrollment task,” says Scott Burke, assistant vice president of undergraduate admissions at Georgia State, on the university’s website. “We would have had to hire 10 full-time staff members to handle that volume of messaging without Pounce.”
The university has enhanced and continued to use Pounce. It has also expanded the chatbot’s role to other student success initiatives.
Tailored Instruction Meets Students’ Needs with AI
Universities aren’t just seeing declines in enrollment, they are also dealing with high dropout rates. Today’s college students need learning to be more engaging and personalized. Technology, especially AI, can help with both those issues. AI, fed with and trained by Big Data, can deliver a personalized learning experience, writes AI expert Lasse Rouhiainen in the Harvard Business Review. Professors can gain unique insights into the ways different students learn and provide suggestions on how to customize their teaching methods to their individual needs, notes Rouhiainen.
Further down the road, an AI-powered machine might even be able to read the expressions on students' faces to tell whether they are having trouble understanding lessons, according to Forbes.
Meanwhile, AI is already making learning more engaging on several campuses.
IBM Research and Rensselaer Polytechnic Institute have partnered on a new approach to help students learn Mandarin. They pair an AI-powered assistant with an immersive classroom environment that makes students feel as though they are in restaurant in China — or in a garden or a tai chi class — where they can practice speaking Mandarin with an AI chat agent.
IBM and Rensselaer call the classroom the Cognitive Immersive Room, and it was developed at the Cognitive and Immersive Systems Lab, a research collaboration between the two entities.
Challenges in Scaling Up the Use of AI and Training IT Staff
While such discrete university AI projects have been successful, they also demonstrate that such initiatives are still very much driven by human intelligence.
For instance, Georgia Tech’s Jill, in her earliest version, was fed more than 40,000 posts from online discussion forums to enable her to answer queries and converse with students. Similarly, Georgia State’s success with Pounce didn’t come easily, notes the Education Advisory Board.
“In addition to the cost of the chatbot’s development, Georgia State’s ten-person team of admissions counselors spent months teaching Pounce how to respond accurately to students’ questions, another task added to a demanding workload,” according to EAB.
Aside from the amount of work this takes, it also means these AI tools will only be as good as the data they’re fed. While for small projects this may be feasible to do, the challenge becomes enormous when rolling out campuswide AI tools that need to use massive amounts of student and institutional data. An AI tool to answer any and all student questions needs a “heavy lift” database, notes Timothy M. Renick, vice president for enrollment management and student success at Georgia State, in The Chronicle of Higher Education.
Campuses where AI is being used in nonacademic areas face other challenges. Their experiences suggest that managing and using volumes of data requires staff beyond IT teams to be trained to use data and AI tools.
For instance, the University of Iowa has connected many campus buildings to computer systems that use AI to monitor them for energy efficiency and any problems. That means staff at these facilities need more than mechanical skills; at the very least, they will need to become effective at incorporating computers and data into their workflows in ways they aren’t today. This means they either need to acquire IT skills or the university IT department needs to offer more support for these teams.
“There’s going to be a skills gap that we’re all worried about,” Don Guckert, vice president for facilities management at the University of Iowa, tells The Chronicle of Higher Education.