The second reading this week is:
Bayne S. (2015). Teacherbot: interventions in automated teaching. Teaching in Higher Education, 20(4), pp. 455-467
After having a good old giggle at my Sci-Fi inspired fantasy of a teacherbot, I decided to do a bit of research prior to reading the above-mentioned article. One interesting example I came across was Jill Watson. Jill is an automated teaching assistant developed using artificial intelligence technology. Her creator, Ashok Goel, introduces Jill to a TEDx audience in this video.
Goel is a professor in the School of Interactive Computing at the Georgia Institute of Technology. When Georgia Tech began their online master’s degree programme, Goel and his team offered a popular course on Artificial Intelligence with up to 350 students enrolling each semester. In one semester alone, the team received 10 000 emails from students asking questions about the course. The team estimated that it would take one teacher, working full time, nearly a year to answer them all.
Many of the student’s questions were repetitive in nature and using data from previous course questions, the team were able to categorise the questions and pair suitable answers to each question type. Jill Watson, the AI teaching assistant, was trained to classify questions and respond with an appropriate answer. When she achieved a 97% success rate in answering questions, Jill was unleashed on the student forums.
Participants on Goel’s online-course were situated all over the world, across many different time zones. Often questions were posted on the forums outside of local office hours. Since Jill didn’t keep office hours “she” was available to answer questions around the clock. Jill was able to engage directly and immediately with students. In fact because Jill’s responses were almost instantaneous, she had an in-built delay in response time. This was done to ensure Jill’s non-human identity remained concealed from students.
At the end of the semester when Jill’s AI identity was officially revealed to the students there was an unsurprisingly positive reaction. Jill had successfully helped many of them over the course of the semester and they had witnessed first hand the benefits of AI in an practical situation.
It is also possible that Jill’s unwavering availability and personal attention to students may have helped motivate students on the online-course who were feeling lost out there in cyberspace.
Massive online open courses have made education both accessible and affordable to the masses. One of the limitations of MOOCs is the feeling of being drowned out by all the “noise” in the chat forums and that comments and questions that were submitted to forums were a mere drop in the ocean of posts. This issue has already been raised on the IDEL course and has been discussed at length on our IDEL forums.
Could teacherbots provide that valuable personal connection that MOOCs by their nature are not able to offer? Might these teacherbots improve the notoriously low completion rates? Various reported completion rates, some as low as 4%, average at about 15% per course according to researcher Katy Jordan.
Despite Goel’s assertion that teacherbots will deliver that much needed personal attention, I am left wondering how personal an interaction with an AI bot feels. I am skeptical of using the positive reactions of the students on the AI course as a measure of how all students would react. They are hardly an unbiased sample set from which to draw conclusions.
I wanted some first-hand experience of bot interaction so I downloaded “Woebot” on my phone. Woebot describes itself as “a philosopher, life coach and friend”. Woebot is an AI app that offers individualised CBT coaching. Woebot’s developers wanted to bring mental healthcare to the masses. Currently the app receives more than two million messages each week from over 120 countries.
During testing amongst Stanford students, Woebot users reported decreased feelings of anxiety and isolation after two weeks of talking to Woebot. Even though they were aware that Woebot was not a real person. It will be interesting to see how Woebot helps with my anxiety levels whilst I write my first assignment in 13 years.
The Jill Watson example of an AI bot is an effective example here – although machine learning based bots have been problematic in more open online environments (see, for example Tay (bot) https://en.wikipedia.org/wiki/Tay_(bot) ). There seem to be two main benefits to these bots in education: (a) providing teacher presence at scale and (b) reducing the need for human responses to common student questions. But is there an ethical issue in not informing students before hand that their questions may be answered by a non-human actor?
The question of whether bots may address the ‘problem’ of low completion rates of MOOCs is an interesting one. The problem assumes that MOOCs are to be understood as complete courses to be completed and the certificates reinforce this. However, it may be that the participants (and a lot of the drop-outs are simply people who sign up to a MOOC but never even start it) don’t necessarily think of them as whole courses but rather as learning resources they can use to address a particular need. Most MOOC participants are already well-educated and may have well-developed competences in self-directed learning and may be participating in a MOOC to meet a particular learning objective rather than complete the course. As an anecdotal example, I signed up for a couple of MOOCs to get some instruction on using particular bits of software which required that I participate in about 30% of each MOOC and I ignored the rest. But that is an anecdote and there’s no clear evidence to support this point.
The Woebot is also interesting example – the UK NHS also use a lot of computer based CBT which has a strong evidence base of success (so long as the ten weeks are completed). I hope you don’ feel too anxious about the assignments
Thanks again for the thought-provoking comments. The Woebot experience has actually been pretty enlightening. I have examined some of my own “black and white” thinking about my academic progress so far. I will blog more about this later.
Thanks. I’ve heard positive things about Woebot so will be interested in readings your reflections.