Sign up for Chalkbeat Newark’s free newsletter to get the latest news about the city’s public school system delivered to your inbox.
Newark Public Schools will use a new artificial intelligence tool this fall to help identify students who may be struggling to read, as part of the state’s new literacy plan aimed at boosting early reading skills.
The Newark school board last week approved the use of Amira, an AI-powered literacy tool developed by San Francisco-based company Amira Learning, as a universal screener, an assessment given to students to identify who may be at risk for reading difficulties. Under the state’s new Literacy Framework, New Jersey school districts are required to screen students in grades K-3 starting this fall.
Amira is designed to assess student reading deficiencies and improve reading skills through personalized intervention plans based on individual needs.
Although the use of AI for literacy screening is a growing trend across the country, the state’s working group on student literacy noted that the tool may have difficulty assessing students with accents or complex speech challenges. That could be a concern in a district like Newark, which has more than 11,000 English language learners, with most speaking Spanish or Portuguese, although some speak Arabic, French, Haitian Creole, or other languages.
The board approved a two-year term contract with Amira Learning that will run from July 1 through June 30, 2027, at a cost not to exceed $900,000.
Amira is the only AI screener recommended by the state’s working group, which helped create the Literacy Framework that is meant to boost early reading skills among young learners starting this fall. The Newark school board last week also approved a $109,260 grant to help fund the district’s screener as part of one of two state grants to help school districts implement the new literacy requirements.
Newark is among a growing number of school districts adopting Amira as a literacy screener. The AI tool is currently used in over 2,000 school districts across the country, including in Oklahoma and Colorado, and has been approved by a number of states, including California, Idaho, and Georgia.
Literacy screeners are reading assessments used to quickly identify students who may be at risk for reading challenges and help educators pinpoint students who may need additional support. Most literacy screeners use digital tools to analyze students’ reading skills, but typically don’t include an AI component.
Using AI literacy screeners such as Amira to identify which students need support could help, but Yaacov Petscher, a professor at Florida State University and associate director at the Florida Center for Reading Research, stressed that AI should add to – not replace – teachers’ expertise, particularly in addressing multilingual diversity and instructional needs.
“We’re in a culture of AI where there’s a rush to integrate, but we need evidence of how well the systems are working without biases,” said Petscher, who also works to build reading assessment tools. “We want to think about how to support and augment systems, but not completely replace teachers and their training for delivering core reading curriculum.”
Amira has been piloted in six Newark Public Schools, and initial data shows increased student growth, according to the district’s request for proposals for screeners. Last school year, the district adopted new approaches to teaching phonics, implemented explicit writing strategies, and began training teachers’ knowledge of the science of reading as part of its plan to boost student achievement after spring state test scores showed student difficulties in writing and English language arts.
Paul Brubaker, the district’s communications director, did not respond to questions about piloting Amira Learning, what type of growth students had shown, and why the district decided to use an AI literacy screener.
What is a literacy screener and how is it used?
Literacy screeners are typically given to young students at the beginning of the school year and again during the winter or spring to test their ability to read, sound out words, and pinpoint trouble areas. Research shows that it’s best to identify at-risk readers at an early age to prevent them from falling further behind as they learn new literacy skills.
Using AI to help screen for student reading deficiencies is new, and the level at which they could improve academic performance is still being researched, according to Petscher, the professor at Florida State. Similarly, an independent research review of literacy tools such as Amira found that AI could be beneficial if it shows evidence of accurately measuring reading deficiencies.
The study also reviewed Amira and found that its measure for predicting reading challenges may not be capturing important aspects of early reading development, but may show accuracy in the future if it conducts a study.
“A company will say, we have a screener that screens for dyslexia, but we have to go deeper to understand well, what do they mean by that?” Petscher added.
Amira Learning was founded in 2018, but the work to develop Amira started at Carnegie Mellon in 1993, according to Mark Angel, CEO of Amira Learning. Angel, who has worked in the AI tech industry for more than three decades, took over Carnegie’s research and collaborated with education experts to launch and train Amira in the science of reading. Amira uses machine learning technology to learn from students and adjust to their needs, Angel added.
Last June, the company announced it had merged with Istation, a reading screening tool widely used in Colorado and other states across the country.
As students read aloud, Amira listens for pauses, hesitations, fluency, and other factors, while pinpointing strengths and areas for growth, Angel said. Student recordings and other data are owned and controlled by each school district, Angel added.
Last school year, Amira reached a more than 95% accuracy rate among all student subgroups, and intervention effectiveness matched or exceeded what would be expected from a skilled human tutor, according to a company assessment.
Angel acknowledged that there are biases with AI based on how each system is trained but the goal with Amira is “to correct that systemic bias” by overexposing the system to different speech patterns and accents so Amira is more accurate in understanding students who have been historically under- or over-identified for reading deficiencies, Angel added.
Similarly, Angel said, Amira is not meant to replace the teacher’s knowledge but rather support their work.
“Our belief is that Amira will help teachers to really get back to a workable, worthwhile situation in the classroom, and in ways that the pandemic and other things have kind of robbed them of,” Angel added.
Literacy screeners aren’t a one-size-fits-all, Petscher said, noting that school districts often use multiple screeners to help inform curriculum decisions and identify student reading challenges.
AI systems “don’t seem to be trained or fine-tuned” well enough to account for linguistic diversity and require humans to double-check outcomes created by systems like Amira, Petscher said.
The state’s working group report noted that teachers are encouraged to listen to Amira’s student recordings and rescore tests if needed, which also helps the software learn a student’s speech profile.
In August 2024, Gov. Phil Murphy signed four bills that launched the New Jersey Literacy Framework and called on school districts to implement literacy screenings for students in K-3, create reading intervention plans, and start new training on reading instruction for pre-K-6 staff starting this fall. The bills also established the Office of Learning Equity and Academic Recovery, or LEAR, in the state’s Education Department, which will oversee the new literacy initiatives.
Jessie Gomez is a reporter for Chalkbeat Newark, covering public education in the city. Contact Jessie at jgomez@chalkbeat.org.
