Search for: All records

This project examines the prospect of using AI-generated feedback as suggestions to expedite and enhance human instructors’ feedback provision. In particular, we focus on understanding the teaching assistants’ perspectives on the quality of AI-generated feedback and how they may or may not utilize AI feedback in their own workflows. We situate our work in a foundational college Economics class, which has frequent short essay assignments. We developed an LLM-powered feedback engine that generates feedback on students’ essays based on grading rubrics used by the teaching assistants (TAs). To ensure that TAs can meaningfully critique and engage with the AI feedback, we had them complete their regular grading jobs. For a randomly selected set of essays that they had graded, we used our feedback engine to generate feedback and displayed the feedback as in-text comments in a Word document. We then performed think-aloud studies with 5 TAs over 20 1-hour sessions to have them evaluate the AI feedback, contrast the AI feedback with their handwritten feedback, and share how they envision using the AI feedback if they were offered as suggestions. The study highlights the importance of providing detailed rubrics for AI to generate high-quality feedback for knowledge-intensive essays. TAs considered that using AI feedback as suggestions during their grading could expedite grading, enhance consistency, and improve overall feedback quality. We discuss the importance of decomposing the feedback generation task into steps and presenting intermediate results, in order for TAs to use the AI feedback. 

This project examines the prospect of using AI-generated feedback as suggestions to expedite and enhance human instructors’ feedback provision. In particular, we focus on understanding the teaching assistants’ perspectives on the quality of AI-generated feedback and how they may or may not utilize AI feedback in their own workflows. We situate our work in a foundational college Economics class, which has frequent short essay assignments. We developed an LLM-powered feedback engine that generates feedback on students’ essays based on grading rubrics used by the teaching assistants (TAs). To ensure that TAs can meaningfully critique and engage with the AI feedback, we had them complete their regular grading jobs. For a randomly selected set of essays that they had graded, we used our feedback engine to generate feedback and displayed the feedback as in-text comments in a Word document. We then performed think-aloud studies with 5 TAs over 20 1-hour sessions to have them evaluate the AI feedback, contrast the AI feedback with their handwritten feedback, and share how they envision using the AI feedback if they were offered as suggestions. The study highlights the importance of providing detailed rubrics for AI to generate high-quality feedback for knowledge-intensive essays. TAs considered that using AI feedback as suggestions during their grading could expedite grading, enhance consistency, and improve overall feedback quality. We discuss the importance of decomposing the feedback generation task into steps and presenting intermediate results, in order for TAs to use the AI feedback. 

We developed and implemented a framework for examining how molecular assay sensitivity for a viral RNA genome target affects its utility for wastewater-based epidemiology. We applied this framework to digital droplet RT-PCR measurements of SARS-CoV-2 and Pepper Mild Mottle Virus genes in wastewater. Measurements were made using 10 replicate wells which allowed for high assay sensitivity, and therefore enabled detection of SARS-CoV-2 RNA even when COVID-19 incidence rates were relatively low (~10 −5 ). We then used a computational downsampling approach to determine how using fewer replicate wells to measure the wastewater concentration reduced assay sensitivity and how the resultant reduction affected the ability to detect SARS-CoV-2 RNA at various COVID-19 incidence rates. When percent of positive droplets was between 0.024% and 0.5% (as was the case for SARS-CoV-2 genes during the Delta surge), measurements obtained with 3 or more wells were similar to those obtained using 10. When percent of positive droplets was less than 0.024% (as was the case prior to the Delta surge), then 6 or more wells were needed to obtain similar results as those obtained using 10 wells. When COVID-19 incidence rate is low (~ 10 −5 ), as it was before the Delta surge and SARS-CoV-2 gene concentrations are <10 4 cp/g, using 6 wells will yield a detectable concentration 90% of the time. Overall, results support an adaptive approach where assay sensitivity is increased by running 6 or more wells during periods of low SARS-CoV-2 gene concentrations, and 3 or more wells during periods of high SARS-CoV-2 gene concentrations. 

The effect of branches on the linear rheology of entangled wormlike micelle solutions is modeled by tracking the diffusion of micellar material through branch points. The model is equivalent to a Kirchhoff circuit model with the sliding of an entangled branch along an entanglement tube due to the constrained diffusion of micellar material analogous to the flux of current in the Kirchhoff circuit model. When combined with our previous mesoscopic pointer algorithm for linear micelles that can both break and fuse, the model adds a branch sprouting process and therefore enables simulation of the dynamics of structural change and stress relaxation in ensembles of micelle clusters of different topologies. Applying this new model to study the relationships between fluid rheology and microstructure of micelles, our results show that branches change the scaling law exponents for viscosity versus micelle strand length. This contrasts with the long-standing hypothesis that branches affect viscosity and relaxation in the same way that micelle ends do. The model also suggests a process for inferring branching density from salt-dependent linear rheology. This is exemplified by mixed surfactant solutions over a range of salt concentrations with flow properties measured using both mechanical rheometry and diffusing wave spectroscopy (DWS). By elucidating the connection between the branching characteristics, such as strand length and branching density, with the nonmonotonic variation of solution viscosity, the above model provides a powerful new tool to help extract branching information from rheology. 

We present a new upper-limb anthropomorphic dexterous telemanipulation system, the Dexterity Testbed Nexus(DexNex). DexNex is teleoperated by a human user in theOperator Station who controls the Avatar Station to complete temanipulation tasks. The Avatar replicates the upper limbs of a human and is statically mounted to the workspace. Three benchmarking tasks were used: box & blocks, the MinnesotaTurning Test revised form (MTTrf), and a table setting task.Subjects completed the tasks with their natural bodies to provide normative data. Subjects then attempted the same tasks with haptic feedback enabled or disabled. The utility of haptics was computed for four metrics. Haptic feedback improved performance for three of the four metrics (26% increase in Box& Blocks score, 12% increased Table Setting success rate, and 1.3x faster time per success in Table Setting). 

Peroxisomes are key regulators of cellular and metabolic homeostasis. These organelles play important roles in redox metabolism, the oxidation of very-long-chain fatty acids (VLCFAs), and the biosynthesis of ether phospholipids. Given the essential role of peroxisomes in cellular homeostasis, peroxisomal dysfunction has been linked to various pathological conditions, tissue functional decline, and aging. In the past few decades, a variety of cellular signaling and metabolic changes have been reported to be associated with defective peroxisomes, suggesting that many cellular processes and functions depend on peroxisomes. Peroxisomes communicate with other subcellular organelles, such as the nucleus, mitochondria, endoplasmic reticulum (ER), and lysosomes. These inter-organelle communications are highly linked to the key mechanisms by which cells surveil defective peroxisomes and mount adaptive responses to protect them from damages. In this review, we highlight the major cellular changes that accompany peroxisomal dysfunction and peroxisomal inter-organelle communication through membrane contact sites, metabolic signaling, and retrograde signaling. We also discuss the age-related decline of peroxisomal protein import and its role in animal aging and age-related diseases. Unlike other organelle stress response pathways, such as the unfolded protein response (UPR) in the ER and mitochondria, the cellular signaling pathways that mediate stress responses to malfunctioning peroxisomes have not been systematically studied and investigated. Here, we coin these signaling pathways as “peroxisomal stress response pathways”. Understanding peroxisomal stress response pathways and how peroxisomes communicate with other organelles are important and emerging areas of peroxisome research. 

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. 

An advantage of Large Language Models (LLMs) is their contextualization capability – providing different responses based on student inputs like solution strategy or prior discussion, to potentially better engage students than standard feedback. We present a design and evaluation of a proof-of-concept LLM application to offer students dynamic and contextualized feedback. Specifically, we augment an Online Programming Exercise bot for a college-level Cloud Computing course with ChatGPT, which offers students contextualized reflection triggers during a collaborative query optimization task in database design. We demonstrate that LLMs can be used to generate highly situated reflection triggers that incorporate details of the collaborative discussion happening in context. We discuss in depth the exploration of the design space of the triggers and their correspondence with the learning objectives as well as the impact on student learning in a pilot study with 34 students. 

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 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.