When Online Teaching Assistants are Really Robots in Disguise

Many students are taking advantage of online education, yet there are some pitfalls that can hold students back; namely, the lack of face-to-face interaction and assistance from the instructor. Now, artificially intelligent teaching assistants might be able to offer some reprieve to this issue.

You might remember how teaching assistants functioned from your own days at the university. They typically act as an aid to the professor, taking on responsibilities that professors don’t have time for, or simply don’t want to do, like administering tests, handing out homework assignments, and helping students who need assistance. In many cases, teaching assistants have been known to actually teach the class and lead in discussions. If students don’t get enough attention or are finding themselves falling behind, the teaching assistant is usually the first person they would go to for answers.

In order to fill this position for online classrooms, some universities have experimented with artificial intelligence programs called “bots” that can help students, should the need arise. One such bot, Jill Watson, was used as a TA for a class in Knowledge-Based Artificial Intelligence at the Georgia Institute of Technology to a degree of success. Jill helped students understand the core curriculum of the class, and wound up being so helpful that students couldn’t guess that “she” wasn’t a real person.

Online teaching focuses on learning management systems, many of which use a discussion board interface that’s designed to facilitate classroom-like discussion. Students can ask questions, which are answered by TAs or the professor. ComputerWorld reports that the KBAI class that’s offered every semester sees about 10,000 messages in its online forums, all of which need to be addressed. This is clearly too many messages and posts for just the TAs and the professor, so implementing an automated solution was a natural step forward.

Ashok Goel, the instructor of the KBAI class, explains that he and his team of graduate students had trained Jill on nearly 40,000 potential questions that have been asked since the class’s inauguration in the fall of 2014. Students that are invested in the subject matter will naturally ask questions, many of which are the same or similar, which makes AI the perfect way to respond to such inquiries.

At first, Jill had some difficulty working as a TA. Her answers would often be skewed off-topic or taken out of context by keywords. After some tweaking, Jill was capable of answering questions properly about 97 percent of the time. ComputerWorld states: “Initially, the human TAs would upload her successful responses to the students, but by the end of March, Jill didn’t need any assistance. She wrote to the class directly if she was 97 percent positive her answer was correct.”

This recent technology is also being used for many businesses. Utilizing automated “chatbot” solutions, businesses can divert some of their basic support to a carefully programmed AI. This AI is simply taught to look for certain keywords in questions, and then pulls up the appropriate answer to help the customer. In the event that the chatbot’s resources are exhausted and it can’t help the customer, it will drive the customer to a live person for a more hands-on approach. These solutions are reachable for small businesses, too - many of them just require time and careful planning to set up, so if you are tracking the types of questions that often bog down your customer service, you can automate them.

We tend to concentrate on the plus-side of working with recent technologies, but taking an objective approach is arguably the most important part of discussing new solutions. As robots continue to grow more realistic, will the line between humans and machines blur? Can developments such as these be good for society, or do they eliminate that which makes us human? Share your thoughts in the comments, and be sure to subscribe to our blog.