Get important education news and analysis delivered straight to your inbox
When Jaclyn Brown Wright took over as principal of Brewbaker Primary School in Montgomery, Alabama, she knew she needed to figure out a way to boost literacy rates. At Brewbaker, which in 2020 served more than 700 students in pre-K through second grade, nearly 20 percent of her students are English learners and 71 percent are economically disadvantaged. In 2019, a year before Brown Wright was hired, less than 20 percent of students were proficient on the school’s reading assessments, the principal said. Brown Wright knew the stakes were high: In Alabama, students can be held back if they are not reading at grade level by the end of third grade.
Brown Wright turned to something unconventional for help: an artificial intelligence avatar named Amira. Amira is the namesake of an AI reading program that aims to improve reading ability by giving kids a personal literacy assistant and tutor. The program listens to children as they read short stories aloud and tracks several literacy skills, including how well they recognize sight words, their ability to decode words and their vocabulary. Students are then given practice activities that target skills they need to work on. Amira shows up on screen as a calm, somewhat stoic woman with short brown hair and a green sweater; she encourages children to try re-reading certain parts of sentences, sounds out challenging words, provides directions for practice tasks and congratulates children when they finish.
Artificial intelligence has been used for years in education to monitor teaching quality, teach classes, grade assignments and tailor instruction to student ability levels. Now, a small but growing number of programs are attempting to use AI to target reading achievement in the early years — a longstanding struggle for America’s schools. More than 800 schools and 355 districts across the country have already adopted Amira as part of their literacy instruction. The program was designed in part by Houghton Mifflin Harcourt (HMH) and based on reading research and reading science from researchers at several universities including Carnegie Mellon University, Johns Hopkins University and the University of Texas Health Sciences. The program’s designers recommend Amira is used by students for a total of 40 minutes each week over the course of two or three days. This summer, Washington, D.C., offered the program for free to students in grades 2-5 who participated in the district’s reading program. And in Texas, the state’s education agency early on in the pandemic offered an AI-enabled computer-based reading platform, Amplio, to every district in the state for free. More than 400 districts in the state have now started using the product, which is offered in both Spanish and English. Amplio, which is meant to help improve reading skills and serve as a reading intervention program specifically for students with disabilities or delays, uses artificial intelligence and natural language processing, a subfield of AI, to analyze speech and reading. (The program also offers digital therapies for children, including speech-language and dyslexia intervention.)
Educators say early results suggest the technology has potential. At Brewbaker Primary School, after two months of using the program, second grade students doubled the number of words they could correctly read aloud per minute. Teachers also reported that many second graders reached a third-grade level of proficiency. Elsewhere, researchers at Carnegie Mellon University found Amira shows promise in boosting the vocabulary of students. In one study, third grade English language learners who used the program increased their vocabularies more than students who worked one-on-one with a human tutor. For students who speak English as a first language, the program led to as much growth as working with a human tutor. And in an analysis of more than 140 students from one Texas school using Amplio, students in the school’s special education program showed progress in speech- and literacy-related goals set out in their individual education programs (IEPs) after using the program. At the beginning of the 2020-21 school year, less than 20 percent of students were able to perform at the level specified in their IEP goal for articulation when speaking. After seven months, 73 percent of students had met their goal.
The growing number of AI-enabled early reading products comes at a time when many young children have already been plunged into the world of online learning during the coronavirus pandemic. But experts caution there are still elements of teaching a child to read that a computer can’t replicate. “If I’m listening to a child read, I might stop, I might ask a question, I might ask, ‘Can you say that again? Let’s read that sentence again,’” said Heather Schugar, a professor of literacy at West Chester University who studies the role of technology in reading instruction.
Human teachers can make adjustments for young readers who have a speech impediment or dialect that does not affect overall literacy skills, she added. Schugar is wary of whether a computer could do that as well as a human. She suggests that if schools use a computer-based program, especially to assess student reading, they should look at data from those programs in partnership with other data points and not use computer-based assessments in isolation to draw conclusions about a child’s reading ability. “I don’t think that [a computer program] necessarily is a bad idea as a supplement, as an extra,” said Schugar. “I worry about replacing the teacher interaction,” she said.
Matthew Mugo Fields, the general manager of supplemental and intervention solutions with Houghton Mifflin Harcourt, said he sees Amira as an assistant for teachers who often spend hours each year assessing student literacy levels. After students read with Amira for a literacy assessment, teachers receive reports that break down how well each student is doing with a variety of literacy skills, including which words and phonemes a student has mastered or finds challenging. “The best use of technology in the classroom is to extend the teacher … and take some of the … more mundane tasks away from the teacher,” Mugo Fields said. “The teachers can focus on higher order things that really drive outcomes.”
In Alabama, Brown Wright has seen the benefit of having an AI program that can easily be accessed both at school and at home, especially during the pandemic. This year, she said, 34 percent of students at Brewbaker scored proficient on the school’s reading assessment, a 15-percentage point increase from 2019. Brown Wright attributes that growth, which occurred during the pandemic, in part to the power of AI. “I have worked in reading and literacy all over the states of Alabama and Colorado,” said Brown Wright. “I’ve not seen gains at the reading foundation level like that.”
Editor’s note: This story led off this week’s Early Childhood newsletter, which is delivered free to subscribers’ inboxes every other Wednesday with trends and top stories about early learning. Subscribe today!