Amira Learning CEO Personalizes Artificial Intelligence For Literacy Gains

Education, like many sectors across the globe, has found an increasing need to develop technologies that support a new ‘normal’ following the initial and transformational impacts of Covid-19 on teaching and learning. 

Many learning challenges for students remain the same even if the landscape has fundamentally changed from traditional brick-and-mortar schools to digital classrooms and e-learning experiences. The time, for EdTech, to answer has come faster than the sector might have previously forecasted and all eyes are on the results of technology investments that are outpacing years prior. 

While advancements in artificial intelligence (AI) have saturated our digital experiences and impacted consumer behaviors, there is still a struggle to see tangible AI applications in the day-to-day of teaching and learning. 

Maybe that horizon is closer than previously expected.

Amira Learning, and their AI-powered reading tutor for children, recently announced the receipt of a $3 million grant from the U.S. Department of Education’s Institute of Education Sciences (IES). The grant will be used to support the 66% of American children who are not proficient in reading, with a specific focus on children in disadvantaged and underserved communities. 

The effort will rest on the shoulders of over 20 years of research conducted at the AI and Robotics Lab at Carnegie Mellon University in partnership with Amira Learning. Time Magazine named Amira’s iOS app, ‘Amira & the StoryCraft’ one of the 100 Best Inventions of 2021

I sat down with Mark Angel, CEO and co-founder of Amira Learning, to better understand the complexities at play that are actually simplifying learning through authentic and intelligent technology.

Rod Berger:  Could you paint the 10,000-foot view of Amira Learning, but more importantly, why Amira Learning personally resonated with you at this stage of your career?

Mark Angel: Great, well, let’s start with what Amira is. It’s the first reading assistant software that can listen to children from five years old read out loud. Amira can assess the mastery and understanding of the reading skills that that student has, or is acquiring, and use that to report back to the teacher to then deliver direct one-to-one personalized tutoring.

For me personally, I started down the Amira path because I had the privilege of being the CTO at Renaissance Learning, which produces the two most widely used school reading applications in America.

What we learned is that even though those pieces of software were the best we could do at the time, they weren’t really solving the enormous problem we have in this country around getting students to fluency by third grade. 

At the same time, we were seeing these incredible advances around speech recognition and AI, and the capacity for software to really support decisions and produce rich experiences in ways that it never had before. So, we felt like the time was now to couple these enormous advances with the huge gap facing EdTech today.

Berger: As a sector, is it fair to say that education is a late bloomer when it comes to integrating technology? Do you think that’s the case here with AI? 

Angel: In some ways it is and my personal journey is a marker of that. I started building AI on Wall Street to support the multibillion-dollar trading world where every little margin of goodness produces massive economic outcomes. It was compelling to take advantage of smart technology and, yes, fairly predictably, AI has made a march from Wall Street to other sectors where the risk of doing long term societal damage is low and the financial rewards are high.

The middle of my career was centered around the ability to put intelligence into e-commerce and customer support where I think we have established the capacity and capability to perform at high levels of reliability and validity. It’s appropriate that we’re now taking it into education where the stakes are so high, but I don’t think we should be too ashamed that we’ve been reluctant to use our 6-year-olds as guinea pigs. 

The 30-year march of AI now makes it pretty safe to turn our attention to a world where it will do the most good and have the highest impact on everyone.

Berger: Most of us grew up watching the Jetsons or other futuristic portrayals of what we might imagine technology to become. Have those experiences provided a higher level of comfort for AI in our lives currently? 

Angel: AI has been so twisted by science fiction or Hollywood that sometimes it is hard for people to distinguish the reality that we experience every day from the worries about what it could become. I think it’s important for us to recognize that unless you’re out in the wilderness you’re touched by AI multiple times a day. Machine learning and advanced algorithms are behind the scenes whenever you use Amazon, Google, or visit your doctor for tests, or almost any activity that involves software. It is making a lot of the decisions about what you’re seeing and what you’re experiencing.

We should begin with the notion that we’re in an AI-driven world, but also recognize that it’s still early. AI is as profound a change as the internet that continues to evolve daily. We’re really only 6 to 8 years into the world where machine learning has taken its place as a predominant part of our lives. We’re already seeing this – our cars are going to be machine learning driven environments very soon and our business environment is going to go from being AI influenced to AI dominated in short order.

Berger: Let’s move the discussion into classrooms where classic software adoption of new technology requires significant buy-in, especially if teachers have been burned by EdTech in the past. Given AI is so ubiquitous outside of school, do you think we’re developing a better relationship with AI that will ease adoption in our classrooms?

Angel: Dead on! We have seen in the last 18 months one very positive side effect of this incredibly horrendous period we’ve been suffering through [Covid-19]. School districts have made 5 to 10 years of progress in becoming technology enabled and friendly, in the span of one year. 

There’s been an enormous accelerant around the school and teachers’ capacity to absorb technology when forced to go through a sea change in a compressed time frame. We had to make that change under such difficult and negative circumstances. We still need to determine the impact on previous software adoption assumptions.  

One thing is clear, the landscape has been reoriented toward a place where the preparedness for technologies like Amira is far higher than just 24 months ago.

Berger: Let’s talk about transformational technology that can impact a student’s ability to access the world and chart their future trajectory. Take me inside Amira on how that’s communicated and how you’ve seen it play out over time.

Angel: Yeah, it’s a fantastic question. I do want to step back for a second and remind everybody about reading. It has become such an integral part of who we are that we forget many 5–7-year-olds find it profoundly difficult.

Reading is an amazingly unnatural act, unlike verbal skills, which have become foundational to who we are as a species. The capacity to look at symbols on a page, translate them into meaning, and then absorb and comprehend what’s being communicated is unbelievably tough for our brains.

We have to rewire the brain to create fluency, yet some students have a harder path to that rewiring, including those with disorders like dyslexia. For almost every kid, fluency won’t kick in without a big foundation of practice. As technologists, we start by having a lot of respect for the science and the skills that we’re trying to ‘softwarize.’ 

At Amira, we understand that we’re just conduits of the research science that’s been done by psychometricians, neuroscientists, reading scientists, psychologists, and cognitive psychologists. Our job is to understand their work well enough to pass it on to our students and teachers with integrity. 

Recent Research Efforts 

Amira Learning, during the 2020-21 school year, partnered with Savannah-Chatham County Public School System (SCCPSS) and the Consortium for Policy Research in Education at Teachers College, Columbia University (CPRE-TC) to explore the associations between Amira Learning and student literacy development. The results, based on data from 2,305 first and second grade students in 29 schools, consistently found that increased usage of Amira Learning was positively associated with student learning.

“Literacy gains among students who completed 30 or more Amira sessions during the fall-to-winter period were 0.45 standard deviations larger than those of their same-school peers who completed four or fewer sessions. This advantage is equivalent to a shift from the 50th to the 67th percentile,” said Douglas Ready, a Professor of Education and Public Policy and CPRE-TC’s Director. 

“Although we cannot yet make causal claims about Amira Learning’s impact, these early findings are quite promising. We are currently conducting large-scale randomized controlled trials in several other school districts to clearly establish Amira’s effects on student literacy development.”

Berger: In terms of the development process, how has it changed and what should be considered moving forward?

Angel: We need to be humble about the development process and realize how long of a journey we still have. There is a lot of inherent arrogance when you have this magical power to turn nothingness into incredibly engaging, wonderfully entertaining software experiences. It’s all too easy for the smart folks in Silicon Valley to steamroll the capabilities and skills that other scientists and professionals can bring to the party. We sometimes talk glibly about EdTech, but it’s hard to combine genuine understanding (from an education point of view) with the talent and capacity to produce high quality technology. 

We are in the early days of melding the talents of educators and technologists to produce the applications that our students deserve.

Berger: Let’s discuss technology specifically. Software that listens to a child’s reading would seem to have application across many domains beyond the initial and primary objective.

Angel: You’re right, but one of the foundational lessons of smart technology, AI, machine learning, whatever label we want to associate with it, is that generalization can be challenging. We typically solve problems by being hyper-focused on the specialized nature of the problem, respecting boundaries, and focusing our energy on solving it in a deep way. That’s been our philosophy at Amira. As we solve those specialized important problems, we are beginning to see options and pathways for generalization. 

Consequently, when a student makes a mistake, we ask ourselves, what is the right form of help needed to accelerate their comprehension to prevent future errors?

This is the fascinating part for me. What we’ve discovered by thinking about the ‘tutoring selection’ or the ‘instructional selection problem’ is how do we select from all available options to help that child? We’re forging a theory of one-to-one assistance as people learn. 

We’re working with cognitive psychologists to understand what triggers we can employ to help people build the capacity to acquire a learning skill. What software can do is put the finger on the trigger to cause the brain to adjust and evolve to be better at that skill. 

As we know more about this process and we improve Amira, as a tutor, and as an instructor, we’re going to open a lot of possibilities for helping students in numerous ways.

Berger: What are your thoughts on social emotional learning (SEL) and the role Amira does or does not play. As I hear you speak, Mark, I think to myself – those who can provide pinpoint accuracy on child curriculum engagement should have a seat at the SEL table. Is that a fair request?

Angel: It’s fair and when we first started building Amira and partnered with Google, now one of our investors and research partners, an employee told us not to underestimate the emotional issues associated with putting an avatar in front of any person, much less a child. We’ve been learning how truthful and powerful that observation is ever since. The whole force and magic of one-to-one tutoring is the dialogue that happens between the mentor and the learner. That dialogue is intensely emotional for the learner and involves a lot of dependence, trust and interaction. 

We’ve come to believe that every learning experience is deeply social and emotional, and we must understand those elements as we’re creating this interaction between technology, students and teachers. 

Berger: Mark, are you prepared for what may happen when people get the keys to the castle and see the possibility within Amira, realizing the utility and alternative applications? What are some lessons learned when it comes to applying a technology for secondary benefit that maybe it wasn’t built for?

Angel: The first lesson we learned is that there’s a big goal in the educational world today to be data and evidence-driven, that matches the policy dominating the discussion. The federal government and many State Departments of Education introduced certifications and criteria around evidence. Amira is rated as strong, which happens to be the highest category there is for ESSA evidence. This demonstrates proof that the technology or program works. But ultimately, our demonstration of strong evidence is not in the dimension of doing nine other things, it’s focused on helping kindergarten through third graders become better readers. 

I think we should start with the standard that the user of the technology, whether a technology company or school district, should align its programmatic use with the umbrella of evidence that’s been established. I think in that way, we can protect ourselves from this migration of goodness into areas of badness. 

We saw this at Renaissance and every education or EdTech company would say the same – no technology or program for supporting students is good if the implementation strays from the high-fidelity guidelines that have been laid down. We just need to work together across the community to make sure that evidence first and fidelity of implementation second, keep us on the straight and narrow.

Berger: What sense of responsibility do you feel for the work that you are doing for our youngest students?

Angel: Well, my family came from Mexico and struggled to adapt to a new culture, language, and set of skills needed.  However, I was incredibly fortunate to find reading very early in life. We moved around and I changed schools a great deal, but my constant companion was a book at my side. Friends that didn’t go away.

What I saw as I progressed in my career was that so many kids who come from opportunity deficient backgrounds fail to find reading early, creating an incredibly negative impact on their life. It was scary to realize that for so many children’s third grade failure is their destiny from which they’ll likely never recover.

Not just me, but everyone on the Amira team wakes up every day with the idea that we’re at a moment where we can make a big difference. Because of that, we have a huge responsibility to be careful and make sure that we do not give the technology and these incredibly powerful ideas a bad name by being irresponsible or thoughtless about how we’re implementing our work.

One could say it is easier to comprehend the frailties of our world by what we can tangibly see – global water shortages, climate changes, superstorms, and pandemics that transform our lives instantly, regardless of longitude or latitude. A counter viewpoint would be recognizing and addressing changes that are invisible, minimal, or dormant in the minds of our youngest community members.

Mark Angel’s work at Amira Learning may be categorized by artificial intelligence, but the willingness of an immigrant to combat the very gateway that once impeded his own sense of self speaks to the authentic spirit of a company willing to push forward.

Ongoing research with Columbia University and the infusion of IES grant funding will most likely shed light on the horizon of Amira Learning and the impact they can have to make the dormant, vibrant and engaged. 

The conversation has been edited and condensed for clarity.