Eric Westervelt, an educational correspondent for NPR, wrote a very interesting article in the Learning & Tech section entitled "." In the article, he makes points normally found in discussions regarding the use of data to identify at-risk students in time to help them.
Then, he moves on to say something astounding in its obviousness: "A lot of times, the discussion will be about students who might be behind or at-risk, but it's also true for students who are really excelling academically. They also need special kinds of attention." In other words, if we're concerned (as we should be) with the impact of one-size-fits-all education on those students who are at-risk, why don't we use the same models to look for students who are outliers at the other end of the distribution? Those of us who have been in the classroom know the students I'm talking about – the quiet ones that sit in the back of the room, volunteer little, and represent a challenge as you can't remember them until you hand that first A back in class four weeks into the semester. If our role as both teacher and mentor is important to the formation of our students, we are just as responsible for identifying students with special abilities that stay quiet as we are with those who are at-risk. Westervelt provides an interesting look into this topic.
On a closing note, I also want to take a moment and encourage our LEAP team's work with High Impact Practice dissemination and the opportunities provided for the Scholarship of Teaching and Learning. My particular background in machine learning has included application of modeling techniques to discovering heart issues. The same techniques can be applied to the data sources mentioned by Westervelt to provide fairly deep models of student behavior. Here's hoping that deep modeling of student behavior and predicted outcomes will share the same academic acceptance as work done in other fields. It should, as a focus on the Scholarship of Teaching and Learning will help bring to fruition the ideas discussed by Westervelt.