Project Activities
To answer these questions, U-GAIN Reading will investigate how generative AI can best be integrated into an existing reading platform and make meaningful contributions to improving education processes and education outcomes of ELs. Specifically, the center will leverage an existing reading platform that serves over one million students and integrates over 20 years of prior research. U-GAIN Reading will pursue four innovations: two innovations to enhance detection of ELs' speech and engagement, and two innovations to enhance adapting for ELs by generating texts to read and interactive conversations about the texts. U-GAIN Reading will integrate these innovations into interventions for two key levels of student proficiency, pre-reading and early comprehension and will ensure interventions are safe before they are used in classrooms. The outcomes of U-GAIN Reading will be: (a) new knowledge about how to use generative AI to create content that matches each EL student's interests and strengths, enable dialogues about the meaning of content, and adapt to a student's progress and needs; (b) evidence about the benefits to ELs and all students of implementing this knowledge in a scalable reading platform; and (c) national leadership on how to support and sustain ELs' active reading in an age of AI.
Focused program of research
The focused program of research includes conducting three exploratory studies to understand the national context for use of generative AI to improve teaching and learning outcomes for ELs around reading and comprehension, as well as to understand patterns of learning in the existing platform. Informed by the exploratory studies, U-GAIN Reading will enhance the existing platform with four innovations to support EL proficiency in pre-reading and early comprehension skills. They will also conduct three pilot studies to assess the promise of the revised generative AI tool for improving learners' education outcomes.
National leadership and outreach activities
U-GAIN Reading will lead a national leadership program focused on Active Reading with Generative AI for ELs and All Students that integrates issues around digital equity, ethics, and evidence, as well as insights on how to apply the science of reading approach to serve the strengths, preferences, and needs of EL students. Through a coordinated approach that amplifies high-quality research centered on EL students and reading, U-GAIN Reading will implement four activities that interweave to achieve national leadership, including 1) showcasing their research through a website that makes research reports and products available to different audiences; 2) engaging educators who work directly with EL students through social media as well as through meetings and conferences tailored to their interests and needs; 3) sharing insights that are responsive to the ongoing and shifting needs of broad audiences; and 4) building capacity of different stakeholders on how to incorporate their innovations in other learning systems to support diverse groups of students, including EL students.
Structured Abstract
Setting
The focused program of research will take place in Washington, DC, Maryland, and Texas. Each area has many schools with high proportions of ELs as well a commitment to using the reading tutor platform.
Sample
The research will include data from over 10,000 elementary school students (primarily 1st and 3rd grades), with a focus on sampling schools with above-average proportions of EL students.
Research design and methods
For the three exploratory studies, U-GAIN Reading will conduct interviews with students, teachers, and school leaders, analyze existing platform data, and conduct classroom observations. The resulting insights will inform both the center's studies as well as the national leadership activities regarding the national context for use of generative AI to improve teaching and learning outcomes for ELs around pre-reading and comprehension skills. For the pilot studies, U-GAIN Reading will conduct student-level randomized trials that include implementation fidelity measurements and cost analysis. As part of the pilot studies, the preliminary impact study is strongly powered to enable asking and answering questions about main effects, moderators, mediators, and implementation fidelity, thus offering evidence both to confirm the impact of generative AI enhancements for EL and all students and to allow development of refined hypotheses for future studies. In addition, the research team will conduct feedback loop studies to complement the exploratory and pilot studies with additional classroom observations, targeted interviews of students, interviews of teachers, and teacher surveys, as well as analysis of platform data. The feedback loop studies will inform the interpretation of research and the focus of further design iterations. In support of the research, a variety of large language models will be evaluated continually to enable leveraging ongoing performance improvements. The research team will use platform data and machine learning to check for and address algorithmic bias.
Key measures
Key measures include platform-based measures of usage and engagement, platform-based measures of reading progress, and the accountability measures that participating districts use in response to state and federal requirements. Within these measures, scales are available for many aspects of reading, such as overall reading mastery, oral reading fluency, word recognition, and comprehension.
Data analytic strategy
Data analysis strategies vary to fit the design of each study. Overall, the research teams will use data to develop generative AI and machine learning innovations. They will also use data to validate those innovations' usability, performance, stability, and degree of algorithmic bias. U-GAIN Reading will analyze data from classroom observations and interviews qualitatively. In the preliminary impact study, researchers will analyze students' demographic and reading performance data using regressions that account for class-level clustering, blocking, pre-test and student-level characteristics, as well as split-sample regression analyses (for subgroup analysis), and mediational analysis.
Cost analysis strategy
U-GAIN Reading will use an ingredients approach to evaluate costs for the platform "as is" and with added generative AI enhancements. They will also report cost per student.
People and institutions involved
IES program contact(s)
Project contributors
SubAwardee(s)
Products and publications
The products from U-GAIN Reading will include generative AI-enhanced pre-reading and early comprehension interventions that enhance the existing Amira Intelligent Tutoring Systems for EL students in grades 1 and 3. Additional products include peer-reviewed publications, conference presentations, research papers, data sets, free researcher training on machine learning and generative AI, and translational content, such as webpages, blogs, and newsletters.
Publications:
ERIC Citations: Find available citations in ERIC for this award here.
Related projects
Questions about this project?
To answer additional questions about this project or provide feedback, please contact the program officer.
