Have you ever started an online course, such as a MOOC? Have you ever finished one? These are very different questions, given that fewer than 7% of people who enrol on them actually make it to graduation. There are many possible reasons for this, but I would argue that one major factor is that we are currently using technology as a bolt-on to educational systems that we already have. Where technology is used simply as a supplementary source of information on top of those already in place (such as textbooks and a teacher), problems of multi-tasking and cognitive load mean that often it can do more harm than good (as I wrote about here). In places with developed and (relatively) successful education systems such as the UK, therefore, I would strongly advocate that technology is bent to fit the system that is already in place, rather than the other way around. However, this situation is not the norm everywhere and it is for disadvantaged communities around the world that the low MOOC completion rate is so tragic, as it is these communities that such courses were supposed to enfranchise. For those who may not have regular access to formal education systems, then, it is clear that the ‘information delivery’ model is not working. Perhaps we need to find whole new educational systems, in which technology can be used more effectively.
Last week I went to a seminar given by Justine Cassell, from Carnigie Mellon University in Pittsburgh. Cassell builds computerised learning tools, but with a huge difference to the standard, impersonal online course (such as MOOCs). Citing research that student-teacher relationships are among the best predictors of progress in school, she feels that one of the primary reasons why we don’t currently learn well from computers is that they lack the “ineffable quality of rapport”. Based on exhaustive analysis of videos of real-life communication, she creates avatars that can act as peer tutors, but tutors which have realistic communication styles, mannerisms and speech inflections the point where we are able to feel an “interpersonal closeness” to them. Interestingly, this closeness is often reflected in impoliteness. Cassell has found that impoliteness between friends is positively correlated with learning; if we feel comfortable enough to be cheeky then we are also likely to learn more effectively from them. Cassell is building all these findings into her peer tutoring models, creating programs to whom we can relate, be cheeky to, and hopefully learn from.
It’s undoubtedly true that most computer-based education programs currently are entirely lacking in rapport (although one friend told me that after watching a full series of Khan Academy videos, he found he couldn’t adapt to a new tutor having got so used to Saalman Khan). What I also like about the idea behind this system is that it totally breaks with the flawed model of used currently in most-computer-based education. This is not to say that I don’t have reservations about the scheme. The current crop of avatars are based entirely around a peer-tutoring model. Whilst this has been found to be a generally effective educational tool, I have my reservations about it being the sole form of delivery. How will some form of assessment be built into the system, for example, and how might the rapport between the pair be affected by the tutor becoming the examiner? If these tools are to be genuinely transformational for people who have limited access to formal education, then they will have to produce outputs, in the form of assessments and exams, that mirror what other students produce. I also hope that the model would be built not only incorporating the lessons of Cassell’s analyses of interpersonal communication, but also with an understanding of educational best practice. It would be a tragedy, for example, to undermine so much promising work by creating a model predicated on some sort of trendy ‘child-centred’ education model, rather than using research-backed pedagogy.
There are also the obvious logistical concerns about being able to deliver such a computationally-heavy system efficiently to places with potentially limited infrastructure to support it. Still, these are interesting ideas. I don’t see Cassell’s avatars ever replacing teachers, but it’s a sad paradox of our unequal world that some of the places where teachers could potentially do the most good are the places where they are not. If we are able to fill some of these gaps effectively with robotic rapport, then I am all for it.