Is There a First Grade Machine Learning Achievement Gap?


Today was the fifth "To Code and Beyond" - a one day conference hosted at Cornell Technion and once again Diane Levitt put together a great show. The theme was Computational Thinking and the day consisted of a variety of talks, panels, and activities. I plan on writing about one panel in particular but for today I wanted to address something that came up as a question. One attendee asked a panel about the achievement gap - the fact that when the CS movement got started in NY some of the more innovative and interesting work was being done with some of our most vulnerable students. Recently, he heard about a student who was doing Machine Learning in early elementary school - first grade. The concern he had was that when we started the less affluent populations might have even been pulling ahead but now he's concerned that the gap between have and have nots will continue to grow even in this new field.

I'm not so concerned about the ML gap in the early grades and I still think that we still have an opportunity to lessen the gap between the haves and have nots with CS and more specifically CT.

I was reminded about when I started fencing. I first picked up a foil in 11th grade. Stuy had a senior heavy team that year. I was the only new junior and there were a bunch of new sophomores. The following year, we were the starting lineup. We were raw but we still managed second place in the city. Most of us continued on to fence in USFA events and then in college. I remember one of the first events we went to. It was the under 20 qualifiers for the Junior Olympics National Championships. There were a couple of superstars who recently came over from eastern bloc countries there. Also a bunch of fencers we didn't know but we knew that they'd been fencing for years - since they were 8, 9, 10. "They've been fencing for years. We've only been doing it for one. How can we ever compete?" I remember Jordan saying "Yeah, they've fenced for years but probably a day or two a week, we've been doing it intensely every day for a year and a half." He also noted that when they were in their early years their bodies probably weren't coordinated enough to really develop so much. He was right. The superstars dominated but then we were very much in the mix and in fact took most of the qualifying spots.

I also thought about teaching CS at Stuy. There were always kid who came in to class with prior experience. That was clear on day one and for the first few weeks. By the end of the first semester you woudn't know it based on overall class performance. The newcomers caught up pretty quickly.

This is why I' not really concerned when I hear about a school doing Machine Learning in first grade. How much can they really do and how deep is it? Is the kid even really ready for it or are they just going through the motions? Or, are they just doing something they've always done in first grade and are just labelling it ML.

I think that last option is fairly likely. It's also why I think we still have an opportunity to close the achievement gap as part of the CS / CT movement.

When people ask me about CS in the early grades I first tell them that I'm not an early grade specialist. I follow this by stating that I think that well resourced schools have been doing thing that we now consider CT for years - logic, problem solving, puzzles, etc.When a group of students solve a problem and describe a solution that includes repetition it used to be just problem solving. Now it's computation thinking. Sure, we can tease it out, formalize it and expand on it but the schools and teachers with the resources and freedom have already been doing this.

With the CS For All movement, we have an opportunity to bring this to all our students not just the well to do ones.

Of course this is easier said than done. We have to prepare them to identify these CT practices and actively develop these skills in their students. This means giving teachers the time, resources, and training and not shorten the already tight tether to standardized multiple choice tests.

At times I've been disappointed with the overall K12 CS Ed movement. We've made great strides coming from nowhere to where we are but as a community we missed the opportunity to bypass the mistakes and bad ideas that teachers of other disciplines are plagued with. We still have this chance with Computational Thinking and the early grades.

I hope we do it right.