Publications
Contextual multi-armed bandits have previously been used to personalize student support messages given to learners by supplying a model with relevant context about the user, problem, and available student supports. In this work, we propose using careful feature selection with relevant domain knowledge to improve the quality of student support recommendations. By providing Bayesian Knowledge Tracing mastery estimates to a contextual multi-armed bandit as user-level context in a simulated environment, we demonstrate that using domain knowledge to engineer contextual features results in higher average cumulative reward, and significant improvement over randomly selecting student supports. The data used to simulate sequential recommendations are available at https: //osf.io/sfyzv/? view_only=351fb8781d2c4f3bbc9d7486762d563a.
Community teachers, particularly those who are Black and Latinx, are assumed to improve retention and outcomes depending on retention in schools that serve low-income Black and Latinx students. Based on a critical quantitative analysis of data collected on the career trajectories and retention of hundreds of alternatively certified mathematics teachers, the study shows that community insiders exhibit significantly higher rates of retention in district schools than community outsiders and, in particular, those from elite colleges. Utilizing quantitative critical theory methodology, the study helps to move the field beyond race-neutral analyses of teachers’ retention and careers. © 2023 AERA.
Knowing how policy-induced salary schedule changes affect teacher recruitment and retention will significantly advance our understanding of how resources matter for K-12 student learning. This study sheds light on this issue by estimating how legislative funding changes in Washington state in 2018-2019-induced by the McCleary court-ordered reform-affected teacher salaries and labor market outcomes. By embedding a simulated instrumental variables approach in a mixed-methods design, we observed that local collective bargaining negotiations directed new state funding allocations to substantially increase certificated base salaries, particularly for senior teachers with 16 years or more of teaching experience. Variability in political power, priorities, and interests of both districts and unions led to greater heterogeneity in teacher salary schedules. Suggestive evidence shows that state average teacher turnover rate was significantly reduced in the first year of reform. The McCleary-induced salary increase particularly reduces mid-career teachers' (8-15 years of teaching experience) mobility rate and late-career teachers' (23+ years of teaching experience) leaving rate. The McCleary-induced base salary increase has mostly null effects on teacher hiring in the first 2 years of implementation.
Limited literature has investigated the effects of state and institutional merit-based financial aid on student choice of science, technology, engineering, and mathematics (STEM) major fields, an unintended consequence with important implications. By leveraging nationally representative longitudinal data from the Beginning Postsecondary Students, we examined these effects-respectively and jointly-with logistic regressions (LR) and propensity score matching (PSM). Both the LR and PSM results showed that students who receive both state- and institution-awarded merit aid were more likely to major in STEM. For students who only received state-awarded merit aid, the PSM presented significant and positive effects while the LR did not. Institution-only merit aid had no statistically measurable effect. We discuss implications for research, policy, and practice for state- and institution-based financial aid.
Single-cell technologies are revolutionizing the entire field of biology. The large volumes of data generated by single-cell technologies are high dimensional, sparse, and heterogeneous and have complicated dependency structures, making analyses using conventional machine learning approaches challenging and impractical. In tackling these challenges, deep learning often demonstrates superior performance compared to traditional machine learning methods. In this work, we give a comprehensive survey on deep learning in single-cell analysis. We first introduce background on single-cell technologies and their development, as well as fundamental concepts of deep learning including the most popular deep architectures. We present an overview of the single-cell analytic pipeline pursued in research applications while noting divergences due to data sources or specific applications. We then review seven popular tasks spanning different stages of the singlecell analysis pipeline, including multimodal integration, imputation, clustering, spatial domain identification, cell-type deconvolution, cell segmentation, and cell-type annotation. Under each task, we describe the most recent developments in classical and deep learning methods and discuss their advantages and disadvantages. Deep learning tools and benchmark datasets are also summarized for each task. Finally, we discuss the future directions and the most recent challenges. This survey will serve as a reference for biologists and computer scientists, encouraging collaborations.
Children’s math performance is strongly correlated with later life outcomes, but early gaps in math skills are stubbornly difficult to close. It is therefore important to identify student math needs early. Using Grade 1–3 student records from Kentucky public schools, the study finds that typically recommended cut scores for widely used early grade math screeners severely under-identify student needs in math. Using optimal cut scores estimated by the Classification and Regression Tree (CART) analysis, the likelihood of under identification of at-risk students decreases by an average of 16 percentage points and sensitivity improves by 28 percentage points. These improvements can be achieved without having to collect new data or administer new assessments.
Interpersonal skills are essential to engineers’ work as they regularly manage large projects, engage with people, collaborate in teams, and offer vital services to society. This study explores how twelve undergraduate students from migratory/seasonal farmworker (MSFW) backgrounds use their funds of knowledge in the form of interpersonal skills in engineering. An hour-long semi-structured interview protocol was designed and used to understand the funds of knowledge participants brought into their respective engineering classrooms. Using thematic analysis, we found that MSFW students accumulated multiple interpersonal skills, such as teamwork, organizational skills, empathetic practices, work ethic, and communication. This work adds to the existing literature on funds of knowledge and puts MSFW students and their unique assets in engineering at the forefront.
As online learning becomes increasingly ubiquitous, a key challenge is maintaining learners' sustained attention. Using eye-tracking, together with observing and interviewing learners, we can characterize both 1) whether they are looking at their learning materials, and 2) whether they are thinking about them. Critically, eye-tracking only speaks to the first distinction, not the second. To overcome this limitation, we supplemented eye-tracking with an egocentric camera, a webcam, a retrospective recall, and mind-wandering probes to capture a 2x2 matrix of attentional/cognitive states. We then categorized N=101 learners' attentional/cognitive states while they completed a multimedia physics module. This meets two goals: 1) allowing basic research to understand the relationship between attentional/cognitive states and behavioral outcomes; and 2) facilitating applied research by generating rich ground truth for future use in training machine learning to categorize this 2x2 set of attentional states, for which eye-tracking is necessary, but not sufficient.
We conducted a 2 x 2 study comparing the digital learning game Decimal Point to a comparable non-game tutor with or without self-explanation prompting. We expected to replicate previous studies showing the game improved learning compared to the non-game tutor, and that self-explanation prompting would enhance learning across platforms. Additionally, prior research with Decimal Point suggested that self-explanation was driving gender differences in which girls learned more than boys. To better understand these effects, we manipulated the presence of self-explanation prompts and incorporated a multidimensional gender measure. We hypothesized that girls and students with stronger feminine-typed characteristics would learn more than boys and students with stronger masculine-typed characteristics in the game with self-explanation condition, but not in the game without self-explanation or in the non-game conditions. Results showed no advantage for the game over the non-game or for including self-explanation, but an analysis of hint usage indicated that students in the game conditions used (and abused) hints more than in the non-game conditions, which in turn was associated with worse learning outcomes. When we controlled for hint use, students in the game conditions learned more than students in the non-game tutor. We replicated a gender effect favoring boys and students with masculine-typed characteristics on the pretest, but there were no gender differences on the posttests. Finally, results indicated that the multidimensional framework explained variance in pretest performance better than a binary gender measure, adding further evidence that this framework may be a more effective, inclusive approach to understanding gender effects in game-based learning.
This work investigates how tutoring discourse interacts with students' proximal knowledge to explain and predict students' learning outcomes. Our work is conducted in the context of high-dosage human tutoring where 9th-grade students (N = 1080) attended small group tutorials and individually practiced problems on an Intelligent Tutoring System (ITS). We analyzed whether tutors' talk moves and students' performance on the ITS predicted scores on math learning assessments. We trained Random Forest Classifiers (RFCs) to distinguish high and low assessment scores based on tutor talk moves, student's ITS performance metrics, and their combination. A decision tree was extracted from each RFC to yield an interpretable model. We found AUCs of 0.63 for talk moves, 0.66 for ITS, and 0.77 for their combination, suggesting interactivity among the two feature sources. Specifically, the best decision tree emerged from combining the tutor talk moves that encouraged rigorous thinking and students' ITS mastery. In essence, tutor talk that encouraged mathematical reasoning predicted achievement for students who demonstrated high mastery on the ITS, whereas tutors' revoicing of students' mathematical ideas and contributions was predictive for students with low ITS mastery. Implications for practice are discussed.


