SRI Education develops and evaluates the effectiveness of programs and tools that leverage artificial intelligence (AI) to improve learning experiences and outcomes for preK–16 students. These innovations are not intended to replace adults, but rather to strengthen the connections between children, adults, and learning content. Below, we highlight a range of projects underway.
Personalizing and adapting learning with AI
- My Reading Academy is an adaptive learning system by Age of Learning that combines reading supports for young children, families, and teachers to support literacy in school and at home. Teachers can use a dashboard to track students’ usage, skill mastery, and areas of struggle and identify ways to support students’ specific learning needs in the classroom. Researchers at SRI and Age of Learning are conducting a quasi-experimental evaluation of My Reading Academy in prekindergarten and kindergarten classrooms. The study will measure whether students in classrooms that use My Reading Academy perform better on end-of-year standardized literacy assessments than students in classrooms that do not use My Reading Academy, after controlling for baseline differences.
- Reading Together: Building Family Literacy Through AI-Enabled Tutoring. Reading Together is funded by a Transformative Research in the Education Sciences grant, administered by the Institute of Education Sciences of the U. S. Department of Education. The project is a 3-year partnership with Amira Learning to develop and test the impact of a scalable, personalized early literacy intervention. In addition to supporting children’s literacy skills development, Reading Together will support adult literacy skills for parents who struggle with reading through eLearning mini-courses delivered as three-way sessions between students, their parents, and the intelligent tutoring system. The three study phases include (1) engaging in a co-design process with parents and teachers to provide Amira with data to create a Read Together prototype, (2) conducting two short-cycle feasibility studies of the intervention prototype and sharing data with Amira to inform Reading Together refinements, and (3) conducting an experimental evaluation of the intervention with 500 kindergarten children, their parents, and teachers.
- Sound Town is a web application from Hogalit designed to support early literacy skills for children in prekindergarten and kindergarten, with a particular focus on building phonemic awareness. Sound Town includes a unique link-sharing feature for teachers to include literacy activities in their regular communications with parents. The application is designed for parents to engage in one individualized 5-minute activity a day with their children. Data from children’s use of Sound Town at home is then sent to their teachers’ reporting dashboard, completing the home-to-school connection.
In partnership with the Sound Town developer, SRI researchers will examine the usability of Sound Town and the feasibility of implementing it in classrooms and homes, which will inform refinements to the application. We will independently conduct a randomized controlled trial of the prototype to examine the impacts of Sound Town on children’s early literacy skills, examine how closely teachers and parents implement the application as intended, and estimate the cost of implementation. Funding for the study comes from a Small Business Innovation Research Phase II Grant, administered by the Institute of Education Sciences. - Yixue/Squirrel AI–Adaptive Learning System: Yixue/Squirrel AI aims to provide high-quality, personalized tutoring experiences that are affordable for K–12 students around the world. Because human-delivered tutoring has been the norm, complementing tutors with adaptive AI requires evidence to build customer trust. SRI Education has worked with Yixue/Squirrel AI to conduct a series of rapid, rigorous evaluations of its AI tutoring system—demonstrating efficacy that leads to market acceptance of lower cost/high-quality tutoring. Through this work, SRI researchers have examined impacts on students’ learning and engagement as well as factors associated with better outcomes. A recent publication in Frontiers in Psychology shared some of the findings about students’ motivation when using an adaptive learning system.
- TROVE: SRI Education recently won a National Science Foundation EAGER grant for Technology to Review Online Videos for Education (TROVE). This project will develop a machine-learning-based tool to identify early childhood mathematics content in online videos that young children watch. This technology will enable new approaches to involving parents in their children’s media use, by helping parents understand the content, and not merely the quantity or titles, of videos their children watch online. The system will automatically classify educational content in online videos and provide tailored, sequenced video recommendations based on learning standards. With internal funding, SRI researchers are also developing a complementary system that applies similar algorithms to identify standards-aligned early literacy content in online videos.
Using AI to strengthen human-to-human interactions
- Automated Collaboration Assessment. Through internal funding and a National Science Foundation Cyberlearning grant, SRI researchers are using our existing vision and AI capabilities to develop a tool that assesses collaboration in real-time with nonobtrusive, video-based monitoring to detect specific nonverbal behaviors associated with effective collaboration. This system will provide automated analytics in nonburdensome, accurate, and timely feedback to students and teachers on specific behaviors considered important for effective collaboration. Additional information about the projects and tool is available in a recent article and video.
- SMARTalk Speech Metrics for Advancing Rich Talk. SMARTalk assesses the quality of child-directed adult speech in early care and education settings. SMARTalk uses SRI’s Speech Technology and Research (STAR) Laboratory innovations in automatic speech recognition, natural language processing, and analysis of the emotional components of speech to provide teachers and coaches with feedback on specific aspects of teacher language considered important for effective instruction.
Still in development, SMARTalk will soon be able to capture information on a larger number of valuable linguistic features than other automated language apps capture. Adult–child speech is usually tracked by either manually rating in-person or video observations or using automated approaches that do little more than count words. SMARTalk will provide more frequent and timely feedback at a lower cost and burden than other tools, which makes it more scalable and more useful for improving teacher practice.
Analyzing learning and test-taking behaviors
- Analysis of NAEP Mathematics Process, Outcome, and Survey Data to Understand Test-Taking Behavior and Mathematics Performance of Learners with Disabilities. Funded through a research grant from the National Center for Special Education Research in the Institute of Education Sciences, this project uses 2017 National Assessment of Educational Progress (NAEP) grade 8 mathematics assessment data to generate evidence that improves our understanding of the link between test-taking behavior and mathematics performance for learners with disabilities. SRI researchers are using state-of-the-art machine-learning techniques to extract action features and time features from the NAEP grade 8 math assessment process. Results are expected to improve the future development and administration of digital learning assessments, identify needed enhancements to mathematics instruction, and highlight areas where further research is needed.
The 2017 NAEP assessments collected data on the outcome of each task and on actions that reflect students’ cognitive processes while they complete each task. Using computer-generated log file data as well as student, teacher, and school surveys, SRI researchers will examine how test-taking behaviors (underlying math cognitive processes, time on task, levels of engagement, and use of accommodations and accessibility supports) differ by disability status and are associated with performance on NAEP mathematics assessment items. We will also investigate differences in the association between test-taking behaviors and test performance for learners with disabilities compared to their peers without disabilities.
