Precision Learning – Umm What?

What does “precision learning” even mean?  I think the first time I heard “precision” used was in relation to “precision medicine”.  That was a term tied to pivoting treatment for patients based on their specific needs based an extensive data analysis.  There are many factors that can impact treatment such as gender, size, family history, and a massive list of other attributes.  It also seems there’s a trend toward a big data project when you throw in the term “precision”. 

In education, everything is focused on a student’s outcome.  So in precision learning, you’d be pivoting to focus on the student and what behaviors help achieve the desired success criteria like grades, test scores, etc.  That’s a tough nut to crack from a data perspective though right?  It is one of those times where I have to admit, this seemed a far away fuzzy concept.  The data capture of data for that seemed beyond this data gal’s realm of thinking.  In a recent project, I found the most interesting data set that immediately brought me to thinking of the term “precision learning”. 

I worked with a customer at a university that had an e-textbook and learning system in place called VitalSource.  There are extracts from this tool that are made available in a few methods that contain all the telemetry of all interactivity from within the application.  So imagine now imagine having the ability to analyze a student’s interactivity with the learning text!  When I learned what data I was to work with on this POC, the data nerd in me felt like I won the lottery!

After a few pinches to myself for landing into this project/data, I did get brought back down to earth a notch when I heard the data came in gzipped nested json files.  I realized I was staring down the barrel of a full on big data project!  I don’t get many of these in education so I was insanely excited.  And I’m even more excited to share what I learned with everyone, especially my friends in the education data world.  I intend to do deeper dive posts into this data and the architecture below.  I also created scripts you can use at https://aka.ms/PrecisionLearning to setup the entire solution and sample data (pulled from sample data posted on VitalSource documentation).  Please comment or reach out if this solution is helpful to you or you have suggestions to make it better.