Search EdWorkingPapers
Students with Learning Differences
Displaying 1 - 10 of 26
Generative AI, particularly Language Models (LMs), has the potential to transform real-world domains with societal impact, particularly where access to experts is limited. For example, in education, training novice educators with expert guidance is important for effectiveness but expensive, creating significant barriers to improving education quality at scale. This challenge disproportionately hurts students from under-served communities, who stand to gain the most from high-quality education and are most likely to be taught by inexperienced educators. We introduce Tutor CoPilot, a novel Human-AI approach that leverages a model of expert thinking to provide expert-like guidance to tutors as they tutor. This study presents the first randomized controlled trial of a Human-AI system in live tutoring, involving 900 tutors and 1,800 K-12 students from historically under-served communities. Following a preregistered analysis plan, we find that students working on mathematics with tutors randomly assigned to have access to Tutor CoPilot are 4 percentage points (p.p.) more likely to master topics (p<0.01). Notably, students of lower-rated tutors experienced the greatest benefit, improving mastery by 9 p.p. relative to the control group. We find that Tutor CoPilot costs only $20 per-tutor annually, based on the tutors’ usage during the study. We analyze 550,000+ messages using classifiers to identify pedagogical strategies, and find that tutors with access to Tutor CoPilot are more likely to use strategies that foster student understanding (e.g., asking guiding questions) and less likely to give away the answer to the student, aligning with high-quality teaching practices. Tutor interviews qualitatively highlight how Tutor CoPilot’s guidance helps them to respond to student needs, though tutors flag common issues in Tutor CoPilot, such as generating suggestions that are not grade-level appropriate. Altogether, our study of Tutor CoPilot demonstrates how Human-AI systems can scale expertise in real-world domains, bridge gaps in skills and create a future where high-quality education is accessible to all students.
Understanding the factors that influence student outcomes is crucial for both parents and schools when designing effective educational strategies. This paper explores the impact of peer age on both cognitive and non-cognitive outcomes using a randomized sample of middle school students. By analyzing how exogenous variations in peer age affect students' academic performance, self-expectations and confidence, health perceptions, behavioral traits, and social development, we highlight the important role that peer age plays in educational contexts. Our findings reveal that an increase in the average age of classmates results in negative effects on both cognitive and non-cognitive outcomes of a student. We also identify significant heterogeneous effects based on student relative age and gender. We delve into potential mechanisms behind these effects and study inputs from the perspective of student themselves, parents, teachers, and the school within the framework of the education production function. The results suggest that students' persistence in their studies, the quality of friendships, and the school environment they are exposed to are the primary drivers of our main findings. These findings underscore the importance of addressing age disparities within classrooms to enhance students' cognitive and non-cognitive development.
Despite well-designed curriculum materials, teachers often face challenges in their implementation due to diverse classroom needs. This paper investigates whether Large Language Models (LLMs) can support middle-school math teachers by helping create high-quality curriculum scaffolds, which we define as the adaptations and supplements teachers employ to ensure all students can access and engage with the curriculum. Through Cognitive Task Analysis with expert teachers, we identify a three-stage process for curriculum scaffolding: observation, strategy formulation, and implementation. We incorporate these insights into three LLM approaches to create warmup tasks that activate background knowledge. The best-performing approach, which provides the model with the original curriculum materials and an expert-informed prompt, generates warmups that are rated significantly higher than warmups created by expert teachers in terms of alignment to learning objectives, accessibility to students working below grade level, and teacher preference. This research demonstrates the potential of LLMs to support teachers in creating effective scaffolds and provides a methodology for developing AI-driven educational tools.
The United States is facing growing teacher shortages that may disproportionately affecting schools serving high proportions of students of color, low-income students, and those in rural or urban areas. Special education teachers (SETs) are particularly in demand. Each year, nearly half of all vacancies are filled with teachers switching from one school to another, yet little research has addressed the nuances of within-career sorting, especially by subject. Utilizing longitudinal data covering 27 years and over 1.2 million teachers in Texas, this study examines SET switching patterns relative to core subject teachers, utilizing discrete time hazard modeling, fixed-effect regressions, and geographic information system mapping. Results show SETs switch schools at much higher rates, associated with experience, salary, and student demographics, yet generally transfer shorter distances than their peers. These findings highlight differential subject-specific labor market dynamics, suggesting targeted recruitment and retention strategies to address widespread shortages.
Currently 15 percent of U.S. students receive special education services, a widespread intensive intervention with variable effects on students. Spurred by changes in federal policy, many states and districts have begun adopting the Response to Intervention (RTI) approach to identifying students to receive special education services. RTI seeks to provide a system for targeting interventions to children facing early academic challenges and identifying children with specific learning disabilities (SLD). This paper uses a difference-in-differences design to examine the effects of RTI adoption across Oregon on elementary students’ disability identification and state-standardized achievement test scores. RTI adoption reduced special education identification by 1.4 percentage points (11%) and SLD identification by 0.5 percentage points (15%). RTI also caused moderately large reading test score gains for Black students (0.15 SD) and did not reduce other students’ achievement. These findings suggest RTI is a promising approach to supporting struggling students.
This study investigates the impact of states' adoption of Response to Interventions (RTI) on the identification and placement of students in special education. RTI, adopted by the reauthorization of the Individuals with Disabilities Education Act in 2004, is designed to improve the identification and support of children with learning disabilities within inclusive educational settings. Using multiple national datasets, we employed a difference-in-differences method to assess state-level impacts from 2004 to 2018. Results show that states adopting RTI observed increased identification of students with specific learning disabilities, yet showed reductions in the placement of students with disabilities in separate school settings. Furthermore, our subgroup analyses suggest that RTI adoption disproportionately increased disability identification among non-White students relative to their White peers.
Students with disabilities represent 15% of U.S. public school students. Individualized Education Programs (IEPs) inform how students with disabilities experience education. Very little is known about the aspects of IEPs as they are historically paper-based forms. In this study, we develop a coding taxonomy to categorize IEP goals into 10 subjects and 40 skills. We apply the taxonomy to digital IEP records for an entire state to understand the variety of IEP goal subjects and skills prescribed to students with different disabilities. This study highlights the utility of studying digital IEP records for informing practice and policy.
This study examines the impact of special education on academic and behavioral outcomes for students with learning disabilities (LD) by using statewide Indiana data covering kindergarten through eighth grade. The results from student fixed effects models show that special education services improve achievement in math and English Language Arts, but they also increase suspensions and absences for students with LD. These effects vary across student subgroups, including gender, race/ethnicity, eligibility for free or reduced-price lunch, and English language learner status. The findings reveal both the significant benefits and unintended consequences of special education services for students with LD, highlighting the complex dynamics and varying effects of special education.
Using a natural experiment which randomized class times to students, this study reveals that enrolling in early morning classes lowers students' course grades and the likelihood of future STEM course enrollment. There is a 79% reduction in pursuing the corresponding major and a 26% rise in choosing a lower-earning major, predominantly influenced by early morning STEM classes. To understand the mechanism, I conducted a survey of undergraduate students enrolled in an introductory course, some of whom were assigned to a 7:30 AM section. I find evidence of a decrease in human capital accumulation and learning quality for early morning sections.