Algorithms, Omelets, and Student Placement
I’ve been sitting here at breakfast pondering a real issue. Which came first, the algorithm or the omelet?
Trust me, these two real issues are related to each other. First, how do we assure that students attend the schools and classes best suited to their needs? Second, who determined that the eggs in my omelet met the qualifications for human consumption? What makes for a great egg, anyway? After all, it’s not like we weigh the chicken to assess the egg’s quality.
The problem with weighing a chicken to determine the quality of the eggs is that body condition is only one factor in determining potential nutritional value. Very few people would select their breakfast ingredients based on a single determinant.
There are algorithms used for that, I’m sure.
Determining student placement isn’t too different. You can’t rely on a single measurement to guess how successful a student will be. By itself, a high-stakes assessment is no more effective at predicting success than grades or any other consideration. Instead, educators have long advocated for multiple measures to predict student success.
That’s where student placement algorithms come in.
Data mining in education
Algorithms establish a systematic and orderly process for making decisions. These decisions take into consideration a variety of data, mined from repetitive actions and bits of information collected over time.
IIT Kharagpur Professor L Sreenivasa Roa says that data mining is “based on various parameters to predict and analyze whether a student (he/she) will be recruited or not in the campus placement. Predictions are made using machine learning algorithms.” His research has shown that specifically, random tree algorithms have proven 100% accurate in predicting the likelihood of student success.
Effective data mining depends on being able to sift through vast amounts of information to determine viability. That’s not something done easily or quickly, regardless of how much caffeine I start my day with.
Algorithms can be an effective tool for fairly and equitably determining students’ placement at any point in the curriculum.
A word of caution
While data mining is an effective tool for student placement, the collection of this data is not without danger. Any data used in student placement algorithms must be masked to protect the privacy of children. Education partner Google already has been charged with categorizing student emails and browsing habits. Without parameters and monitoring, data mining could steal private information.
Our use of algorithms for making decisions will continue to grow, and it’s unlikely they’ll be given a second thought, certainly no more than you give to the eggs in your breakfast omelet.
As I’ve pointed out before, “driven by algorithms, students will be placed in groups and given assignments. Students will often work collaboratively, broken into groups through a calculation of strengths, weaknesses, and interests.” We need a system like that. It must be reliable, consistent, and equitable in helping to determine the best placement possible for our students.
It’s only when we meet students at their last point of success that we can take them further along their academic journey. It’s algorithms that will help us do that.