Woo-hoo! The BRIGHT podcast is officially back for Season 2!
Over the past few weeks, I’ve been busy interviewing a wide range of educators and innovators, and now my team and I are looking forward to sharing their stories with you every Wednesday for the rest of the summer and early fall.
For our first episode of Season 2, I had the honor of interviewing Lou Aronson, the CEO and founder of a company called Discourse Analytics, which is leveraging artificial intelligence and machine learning to design tools that analyze student behavior and generate suggestions for meeting student needs.
I asked Lou to tell me more about how AI can be used to further student learning and what advice he has for educators who are perhaps feeling a bit skeptical about these new advancements in technology.
Here’s a sneak peek at our conversation:
Nikki: Hi, Lou! It’s great to have you on the BRIGHT podcast. To get started can you tell us a little bit about yourself and what you do at Discourse Analytics?
Lou: I have a somewhat unusual path into being the CEO and founder of an AI-based company that focuses on edtech. I was a practicing lawyer for 20-something odd years, and I did a lot of work in politics.
As part of my work, I had this idea to start a technology company that was a political social network. Our hypothesis was that issues mattered more to voters than demographics or party, that you don’t vote for a particular person more often than not because of your demographics or the party you belong to, but the issue that matters most to you: gun rights, women’s gynecologic health, marriage equality, food, etc.
Even the way we use some of the terms can indicate a preference for certain aspects of those issues. I searched around about the topic and things of that nature for a couple of years while I was running my law firm.
Then, in early 2011, I met my co-founder, Vijay Perincherry, who is an applied mathematician, and Vijay explained to me, the guy who got a one on his AP Calculus exam in high school, how math and applied mathematics — not statistics, but math — could be used to understand how people make decisions around politics.
But what do we do? What is Discourse Analytics is the second part of that question, and, and D.A. — Discourse Analytics is a little bit of a mouthful, so we call it D.A. — is a prescriptive analytics company that uses artificial intelligence and machine learning that has been combined with things like cognitive science and behavioral economics to understand students using their mindsets, predict outcome areas of mastery and progression, and then create nudges or data objects, which are calls to action to help the student progress and overcome obstacles.
Nikki: Maybe we could get concrete. Can you give me an example of what this might look like for a student?
Lou: Most folks, I think, involved in this podcast are familiar with FAFSA verification in higher ed, right? The application process to get a student loan. The problem with FAFSA verification is roughly a third of all students who apply for federal aid get flagged for verification. Well, the process is so difficult, or non-user friendly, that roughly 40% of those people never complete their verification form and wind up dropping out of school because they don’t get the money.
So, we were able to take 17 data points from one of our partner institutions and build attitudinal profiles on each of the students. These students aligned to three primary think alike clusters. We named them. We had the overwhelmed students, the procrastinators, and the overcommitted.
We then delivered targeted messages, so you can already imagine the message to a procrastinator is going to be much different to an overwhelmed student than an overcommitted student. We did an A/B test, and the control group got the standardized message of “you need to fill this stuff out.” The test group got our targeted messaging, including two messages per week for three weeks. Over three weeks, we improved completion — not just click through and open rates — but the actual completion of the verification forms by 23%.
So, if you think about how it would work with a student, let’s say a student is struggling with social emotional learning loss in high school. I know that this is something that’s obviously top of mind for all of us these days, where a student is seemingly socially disconnected.
Maybe that student has a nervous or a capricious or a fixed mindset, so we would be targeting a message back to that student to possibly get more involved in clubs or activities at the school, get more involved in maybe some outward-looking, community-based activities to feel more connected. Your connection is such an important part of the education experience.
If the student is having academic struggles, we could direct them — depending upon their profile — to a self-help portal for tutoring or to go see the teacher for extra help. Imagine a student with a disorganized profile that is having academic problems. We could route her a piece of content from LinkedIn Learn on how to better manage your schedule.
That’s how some of those pieces come into play. You can almost think about it as a demographically unbiased recommendation engine, like everybody experiences with Spotify, Hulu, Netflix, or Amazon.
I think that one of the biggest things that I try to work toward when I talk to folks is demystifying this and explaining that, you know, teaching is hard. It is. I have a lot of friends who are teachers. I spend a lot of time with teachers.
Over the past 20 years, there have been billions of dollars invested in education, but outcomes haven’t changed. How can we allow teachers who have this desire to help people get educated do more?. Part of this is also making tools like ours less scary. It’s not about replacing you, it’s about augmenting you.
Nikki: Especially considering the average number of students in a class is about 30 kids. If you consider that the average teacher has five classes or so, that’s about 150 students, which is a lot for one person. We so often hear that teachers are incredibly overworked and stressed.
Lou: Yeah, and I know this isn’t the case, but let’s just assume that a teacher taught five hours a day, that’s 300 minutes a day, right? 300 minutes. That’s two minutes per student.
People don’t scale, and computers don’t have intuition. But where we’ve started to get now is that AI can allow us to start having intuition to then let people scale.
Nikki: Can you tell me about who your favorite teacher was and why they were your favorite?
Lou: There were three. I couldn’t figure out which one was my favorite. There was Mrs. Kaplan, my AP English teacher in high school. There was Dr. Bach, who was my first-year writing professor at Michigan. Then, there was Professor Lieberthal, who was the head of the Chinese Studies Department at the University of Michigan when I was a student.
What struck me about all three of them was they knew me and they wouldn’t let me cut corners. Because I was a corner cutter.
Nikki: So, if there’d been a Discourse Analytics in your LMS. . .
Lou: Yeah, if there was a synthetic profile for corner cutters. I was a good student, but they knew I could be better, and they challenged me to be better. They each taught me these little different things that stick with me all the time in terms of how to not use passive voice and how to not use indefinite pronoun references and how to be very specific about the stories that I’m telling and how I’m telling them and how I need to build an evidentiary foundation for my hypotheses.
But they would not accept me not doing my best. That was sometimes painful, and it didn’t always result in the best grade. In each one of those situations, I would say they were my favorite because they knew me, and I believed that they cared. That caring connection is one of the things that we see a lot of as being critical to unlocking improved outcomes.
Nikki: What would you say is your vision for student learning? So if it were up to you, what would you want to see for every single student?
Lou: What I would like to see for every student is for the pathways for successful progression to have as many barriers removed from it as possible.
There aren’t many people in this world who are luckier than me. I didn’t grow up in the richest family, but I grew up in a middle class family in a nice neighborhood, right? I never wanted for anything. A’s were expected, but if I wasn’t getting an A, my parents would either sit me down and make me do my homework or get me a tutor.
It wasn’t if I was going to college, it was where I was going to college. It was: What are you going to be when you grow up? A doctor? A lawyer? A business owner? There were never any limiters.
Thinking about food, finances, and access to information, what I would love to be able to see is — and obviously, I think a lot of this because the way I see the world is driven by data — that there’s a framework in place that allows organizations and governmental entities to unlock the the potential of an individual’s data in such a way that can be beneficial to them at each stage along their pathway.
I’d love to see that resources can be made available to them to remove obstacles when and if they appear, assuming they don’t already have resources to remove those obstacles.
Nikki: One thing that really struck me, you know, when you were speaking earlier — especially now that I’ve had time to kind of let it percolate — is what you said about when you were talking about demographic information and how, as humans, demographic information, in particular, may bias us. We may see a student in a particular situation and make some assumptions about their behavior.
You seem to be saying this algorithm — I’m hesitant to use the term unbiased — is at least differently biased, so it has a different set of criteria that it’s using to look through the data and evaluate.
I think that’s pretty fascinating to consider being a teacher and having the own your own stream of information — like you said, through your own intuition and your own experiences and your own filter — and having the ability to at least look at the recommendations coming from this less biased or differently biased system that’s looking for different things.
Lou: I think the one thing I would like to say about how technology facilitates my vision is that technology is only going to be as good as the people who are putting it into practice.
Technology is a tool, and you are still the prime mover and user of that tool. If you put a carpentry hammer in the hand of a carpenter, he’s going to do some great work like you can see in the back of your room there. You put that in my hand, and it’s going to look like Frankenstein’s monster. It’s just not going to do the same thing.
Nikki: If you were talking to a teacher who was looking to dip their toes into the use of AI and data to advance student learning in their classroom, what kind of advice would you give them?
Lou: Don’t be scared. Challenge yourself the way you challenge your kids. If you were a teacher of history, and a kid wasn’t understanding what was taking place around the American Revolution, for example, you’d have to break it down into digestible bites.
What I would say to people is: Do you have Amazon Prime? Do you watch Netflix? Have you ever accepted a recommendation from Amazon or Netflix on a product or show? Have you ever clicked on a song that’s been recommended to you by Spotify?
These are all user experiences that are being driven by data. We won’t talk about Facebook and Instagram and some of the nefarious things around data in the way algorithms are working.
You’re experiencing data and A.I. and technology and in almost every facet of your life. One of Henry Ford’s great lines was “If I’d given my customers what they had wanted, I would have given them a faster horse.” Elon Musk, on the other hand, has created a paradigm shift — whether you like him or not — in how we look at electric cars.
As teachers, you’re sitting here at this really unique moment of time in which you have an entire generation who are being referred to as “digital natives” and are being raised with technology in their hands at every point in time.
Start trying to meet them where they are. Your lesson plans can change. The technology will continue to adapt and get better, and you’ll get more power out of it. The reason you pay $110 a year for Amazon Prime is you’re getting value, and this is how cloud and data and AI and platforms integrated with responsible business process change can really — on a scalable, one-to-one, and equitable basis across both learning and social emotional loss issues — improve outcomes and overall learner progression for a lifetime.