Back to Blog

Alta wouldn’t be possible without a masterful group of Knerds taking care of business behind the scenes to provide you with the best product imaginable. To give you a little peek behind the curtain at this imagination station, we met up with two of the Knerds responsible for Knewton Alta Calculus to get the 411 on their love of Knewton, upcoming trends in education technology, and the key features that make this new Calculus course so impressive.

Andrew Jones Greg Hitt
Andrew Jones is a data scientist and Alta product manager with a Ph.D in Music Theory. From analyzing jazz piano recordings to studying student learning patterns, Andrew focuses his time on using mathematical modeling and machine learning to extract patterns from complex data. The resulting algorithms help power our adaptive engine and provide students with individualized learning experiences to help them master complex subjects like Calculus. We can all thank Andrew for those A’s! Greg Hitt (now in his adult years) is a former calculus instructor turned Knerd, who spends his days developing courseware that meets the challenges he and his students faced day after day in his own classroom. Using that background, he has worked on many of the pedagogical interventions behind the student experience, designing everything with an eye towards the idea of assisting students on their path to mastery – whether that be the highly detailed answer explanations, mapping the relationships between sub-learning objectives, or building interactive graphs to explore and learn. He really does love Calculus that much!


1.   You’ve been working in the Higher Education field for a while now. Give us a little overview of your past experience and why you became a Knerd.

AJ: I did my undergrad work in the Princeton physics department, where I specialized in large, noisy datasets at the intersection of particle physics and cosmology. When wrapping up my senior thesis, I decided I liked the tools more than the subject matter—so I earned a Ph.D in Music Theory from Yale, where I applied similar computational modeling approaches to complex musical datasets. That’s where my interest in machine learning really started. When I started looking for meaningful applications of machine learning on the data science market, Knewton seemed like a great chance to apply the skills I already had, learn a bunch of (k)new ones, and stay close to the classroom. I started on the data science team in summer 2017, during the big run-up to Alta’s commercial launch, and I added a product management role in late 2018.

GH: I worked as a teacher at a public high school in Brooklyn, New York, for 6 years. I primarily taught AP Calculus, Physics, and Algebra 2, but at some point or another, I taught every level of math offered. Along with teaching 130 students over 4 different subjects every day, I also built some tooling to help with a schoolwide data initiative. After exams, I could run reports on how my students did on individual learning objectives to help guide them in where they needed to study or what topics I might need to review in my classroom. I could also gather information on how effective my teaching was to improve lesson plans for the following year.

Once I started thinking about using data in this way, and seeing how assessing in a low-stakes way helped students both in terms of focusing their time and also in meta-cognition, I completely shifted my grading schema to a mastery-based classroom. I didn’t penalize students for not getting the content initially but saw testing as a formative experience. If they missed an objective, it informed both myself and them of deficiencies and maybe highlighted prerequisite knowledge gaps that I could address in tutoring during lunch or after school. It really transformed my classroom, but this was a lot of work—so seeing that Knewton was trying to do this exact thing sold me immediately.

2.    Given your history in Higher Education, what do you think makes Knewton Alta different?

AJ: Alta didn’t start from the assumption that a fixed textbook or set of assessment questions should work for everyone. Knewton had been providing personalized, adaptive experiences based on individual student needs for years, but our publishing partners ultimately controlled the pedagogy, user experience, and how our adaptivity fit into their educational products. With Alta, we had the chance to start from scratch and say: if we have this great adaptive engine, what do the learning science and educational data mining literatures tell us would be the most impactful learning experience for students? We came to the conclusion that implementing mastery-based learning in the lowest-cost, most scalable way possible would let us extend the benefits of one-on-one attention and coaching to a huge number of students who otherwise might not have it. Alta’s built on that bedrock: putting achievement within reach for every individual student, especially in classrooms where students come from all kinds of educational and financial backgrounds.

For those heterogeneous classrooms, alta is really the first product to take seriously the idea that you might implement mastery-based learning in a variety of course structures. We have instructors choose what learning objectives students should get great at and by when, and each adaptive assignment starts all students on those target learning objectives. Our proficiency model and recommender then provide as much support as is needed to get students across the finish line as efficiently as possible — whether that’s just some worked answer explanations, or some detailed instruction, or potentially significant prerequisite support, all delivered seamlessly and just-in-time.

3.    Knewton Alta was recently aquired by Wiley, a boutique publisher who has been in the Education Publishing space for quite some time. How has that affected Knewton Alta as a company and the courses you create?

AJ: You say “boutique,” but I was nervous about joining a big publisher! Knewton was like 100 employees for most of my time there. But the results have been great — the Knewton team has remained focused on Alta, and our data science team is starting to expand the reach of our models into other Wiley products, so we’re kind of like an in-house tech company. The cross-pollination has had a big impact; we immediately gained access to a huge amount of high-quality content, and Wiley has been able to start thinking about investing in adaptivity in a totally new way. We think this will speed up the rate and increase the quality with which we can launch new Alta titles.

4.    Knewton Alta Calculus was just launched last week and is the first course coming out after the Wiley acquisition. Can you tell us a little about this course and what makes it special/why people should pay attention?

AJ: The calculus market has been static for a long time—there’s one market-leading textbook that’s been updated many times, but the pedagogy hasn’t really changed. And nobody has been able to come up with a strong adaptive or mastery-based solution in such a complicated subject. Calculus is amazing and elegant and important to a bunch of majors and jobs, but it remains a course where a lot of college students give up on math. We’re excited about Alta Calculus because it’s built to serve everybody and to be capable of more robust, targeted interventions than any previous calculus courseware. By wedding a huge quantity of OER content (most of which is generated in-house thanks to Wiley’s financial support) to these interventions, we’ve built a pedagogical model that we also think will scale to other complex subject areas in the future.

5.    Adaptive technology is a big part of the mastery-based learning system Alta provides. How would you describe the adaptive model that Alta uses?

AJ: Alta’s adaptivity is built on a proficiency model and a recommendations engine. Our proficiency model assesses student knowledge states in real time—after every question a student answers in Alta Calculus, we’re jointly estimating how proficient the student is on the highly granular problem type they answered, on the learning objective more broadly, and on related learning objectives like pre- and post-requisites. Our recommender then makes use of those estimates and our highly parameterized content to choose the best path forward for the student: you can think of it as an encoded pedagogy, making decisions about what kinds of problem, instruction, or remediation would be most impactful at any given time.

6.    Outside of adaptive learning, what are the key features of Alta that you would like instructors to know about?

GH: In my mind, the biggest highlight other than adaptivity is the use of desmos. Desmos as a company has been so focused on great pedagogy from its inception, so being able to harness what they have done to build both assessments and explorations that help students get a more conceptual understanding of topics is something I am very excited about.

7.    And last question. We know you spend a ton of time dedicated to your work, but when you’re not creating adaptive algorithms or programming modules, what do you like to do?

AJ: Well, the music thing never really went away, so I’m an avid record collector and hi-fi enthusiast. My wife and I also have an energetic golden retriever, Lucy, who runs most of my non-work life.

GH: I’m an avid birder [200 species last year!] and have a weird obsession with visiting as many state capitol buildings as I can [just hit 31 with Denver!], so I try to sneak in as many weekend trips as possible to do one or both of those things. My commutes are filled with books and crosswords, and I both play and host trivia.


Want to check out Knewton Alta Calculus first hand? Schedule a demo!