Improving Girls’ Math Scores with Emotion-Sensitive Tutor
March 15, 2010 at 1:40 pm 4 comments
In earlier studies conducted in Greenfield and other schools, the software has improved student math test scores by 10 percent, a critical difference for those who are struggling to pass. As Woolf explains, “Our original work was to find out where girls needed extra attention and how to give it to them. According to our studies, the extra support they need compared to boys is more about emotion than information.”
The part that I think is really fascinating is how the computer senses emotion.
Most recently, they’ve added sensors and cameras so the computer can recognize when students are happy or stressed, fidgeting, frustrated or feeling confident. Guided by such cues, the “learning companion” character reaches out with encouraging words to praise a student’s effort, offer a hint or suggest that trying again is an important aspect of learning.
The article seems to suggest that it’s not a gender-specific need for more emotion-sensitivity, but that it’s about trying to correct waning interest in girls. The emotion sensitivity is about trying to remind the girls that they used to like math.
Entry filed under: Uncategorized. Tags: BPC, broadening participation in computing.
4 Comments Add your own
Leave a comment
Trackback this post | Subscribe to the comments via RSS Feed
Computing Ed Research - Guzdial's Take 
1.
Raymond Lister | March 15, 2010 at 2:30 pm
On the issue of sensing emotion, there was a really nice paper about this at ICER 2009, describing how the authors went about sensing emotion in novice programmers. Here is the reference to that paper …
Rodrigo, M. T. and Baker, R. S. 2009. Coarse-grained detection of student frustration in an introductory programming course. In Proceedings of the Fifth international Workshop on Computing Education Research Workshop (Berkeley, CA, USA, August 10 – 11, 2009). ICER ’09. ACM, New York, NY, 75-80. DOI= http://doi.acm.org/10.1145/1584322.1584332
and here is the abstract of that paper …
We attempt to automatically detect student frustration, at a coarse-grained level, using measures distilled from student behavior within a learning environment for introductory programming. We find that each student’s average level of frustration across five lab exercises can be detected based on the number of pairs of consecutive compilations with the same edit location, the number of pairs of consecutive compilations with the same error, the average time between compilations and the total number of errors. Attempts to detect frustration at a finer grain-size, identifying individual students’ fluctuations in frustration between labs, were less successful. These results indicate that it is possible to detect frustration at a coarse-grained level, solely from coarse-grained data about students’ behavior within a learning environment.
2.
Gary Litvin | March 15, 2010 at 3:03 pm
An interesting concluding sentence: “remind the girls that they used to like math.” When did they? Perhaps that was the time to teach them math. At what age do girls lose interest in math? Has anyone heard of any research that would demonstarte that boys and girls should start school at the same age?
3.
Alan Kay | March 16, 2010 at 7:25 am
Lots of interesting stuff here.
First, historically, some of the earliest attempts to gather real-time information about what the user is doing go back to Marvin Minsky and Sam Gebner (a Marvin grad student), Nicholas Negroponte and Dick Bolt (a researcher in Nicholas’ ARC-MAC groupt at MIT). There was some pupilometry work which predated this. We did some follow up work using GSR, gaze, and pupilometry when at Atari. There have been ongoing studies for quite a few years in Japan, etc.
The basic idea is that the mouse and keyboard only “watches what your hands are doing when they are doing something” — otherwise, as Nicholas likes to say, “an infrared toilet knows more about what you are doing than your computer”!
Second, careful studies of master teachers (especially Tim Gallwey, the sports teacher) showed that one of his main processes was to read faces for “the start of lack of interest” and to change the suggestions only then. Betty Edwards, the drawing teacher, has determined that you need at least 1 assistant for every 7 learners, for the very same reasons, and to not let anyone go down a really bad path. (It can be argued that the idea of quick correction is at its best with training, and misses “the making of important mistakes” that is good for higher level thinking and solving.)
Third, girls around 10-12 are typically a bit better than boys at both math and computer programming (and as has been known for a long time, with reading and writing, etc.).
But the real questions seem not to be even asked here (and it doesn’t seem clear what they are actually picking up about improvement).
To just pick one, what if the computer avatar could give more than just encouragement? If it could really help, then we have a much more critical UI situation where we don’t want the learners to turn if off, and this means that the human factors of all the interactions has to be done so much better than e.g. this one.
Cheers,
Alan
4.
Ryan Baker | March 17, 2010 at 10:36 am
Thanks for the hat tip, Raymond!
There’s quite a thriving literature in assessing various aspects of the learner in educational software, including both affect and a variety of behaviors associated with learning, and in responding to individual differences in those behaviors, improving learning.
Arroyo and her group have done some key work in adapting to gaming the system, for instance — a behavior associated with poorer learning in many types of learning software. (the student tries to trick problem-based educational software into letting them advance without learning the material, by systematic guessing or mis-using help).
Arroyo, I., Ferguson, K., Johns, J., Dragon, T., Meheranian, H., Fisher, D., Barto, A., Mahadevan, S., Woolf, B.P. (2007) Repairing Disengagement with Non-Invasive Interventions. Proceedings of the 13th International Conference of Artificial Intelligence in Education. IOS Press.
Their software not only reduced gaming the system, it also improved learning.
My group has also done some work on detecting and remediating gaming the system, and on detecting off-task behavior. We were also able to reduce gaming the system and improve learning…
http://users.wpi.edu/~rsbaker/publications.html
And Art Graesser and Sidney D’Mello have also done some great work in detecting student affect in educational software with sensors (and without)…
It’s a very active area, and I believe that it will be making an impact in a lot of areas of education in the next few years, including CS Education.