How can we encourage girls to enjoy maths and science subjects? Is there some way that we can tempt girls across the great geek divide to the point where they consider STEM careers? There are of course lots of possible answers to this. One which I came across recently in an interesting press release is to develop intelligent tutoring systems tailored to improve maths scores among girls. Woolf and her colleagues at UMass Amherst were interested in providing extra support for girls, and decided to include an emotionally sensitive component to the tutoring system which would give encouragement and help based on the students’ detected affective state. As far as I can tell from the press release, the research results from their field studies are not yet available.
The tutoring system is personified with an optional animated agent who offers emotional support and other advice tuned around the users’ emotional state. It turns out that the boys more often turn off the agent than the girls, which is consistent with some of my own previous work where older boys (12 years) were more likely than girls or younger boys (10-11years) to prefer a tutoring system without an animated agent (see Robertson et al. 2004 if you care). Maybe boys don’t want emotional support. Or maybe they do but they don’t want it conveyed through an animated character. Users of the wretched Microsoft Clippy might sympathise with that view point. (“I see you are feeling upset. Would you like me to a) contort myself into a disturbing hug shape b) administer an electrical shock to see if it helps you focus or c) give you an easier maths problem?)
But how on earth do you reliably infer a user’s mood from their behaviour when using the system? This is a fascinating research topic in its own right, quite apart from the issue of how the tutoring system then responds sensibly to this information. In the UMass research, they use a combination of sensors and cameras with which “the computer can recognize when students are happy or stressed, fidgeting, frustrated or feeling confident”. They also use cues such as “the time taken to answer questions; number of hints requested and grip on the mouse.” As you might imagine, there are huge problems in working out which items in this potentially massive volume of data are most indicative of the users’ emotional state. There is also a reliability problem; Woolf and colleagues report 70 – 80% accuracy when using the camera. Is this reliable enough to base a tutoring system action on? What are the repercussions when it gets it wrong? These are interesting open questions.
I recently heard a talk by Madeline Balaam which had some intriguing thoughts about how teachers might respond to learners’ emotional states (Balaam et al in press). She made the point that researchers have hereto assumed that human teachers know how to respond to their pupils’ feelings assuming they have the information to diagnose them. This turns out to be not necessarily true. In Madeline’s research, she designed Subtle Stones which are tangible devices which high school students can use to convey their emotional state by squeezing the stone to change its colour. The colour/emotion mappings are private to the student (blue might mean happy for me but bored for you) but the teacher has a special display which translates the individual emotion mappings to a standard display, meaning that she can get an overview of the emotional state of every class member in real time. In contrast to Woolf’s system, the users self-report their emotional states rather than having the system infer them from behavioural cues.
I have to admit that I would find it challenging to know what to do if I had access to such a rich source of information about the feelings of my students (beyond the obvious few who fall asleep on the desks). Are they unhappy because they split up with their girlfriend or because I bore them to tears? Is it OK for a learner to feel frustrated for a time if they eventually indicate that they are feeling proud because their work has paid off? I can imagine a first year programming lab class where everyone spends most of the class frustrated and then there are little bursts of pride as people finally get their compiler errors sorted out. What do I do if half the class is happy but the other half is bored? (This is presumably not a problem for an ITS which is dealing with one student at a time.)
To conclude: Woolf’s project is a great example of research which tries to solve a real world problem by tackling extremely challenging research issues. As this particular real world problem has an impact on girls’ attainment and interest in maths and sciences, various people in the ACM community will be watching with interest.
Balaam, M., Luckin, R., Good, J. (In press). Alsmeyer, M., Luckin, R., Good, J. Jump Starting Affective Communication in the Classroom with the Subtle Stone. International Journal of Learning Technologies.
Robertson, J., Cross, B., MacLeod, H., and Wiemer-Hastings, P. ( 2004) Children's interactions with animated agents in an Intelligent Tutoring System. International Journal of Artificial Intelligence in Education. 14.335-357.
No entries found