arXiv subjects

Renzhe Yu

Publications and source records attributed to Renzhe Yu.

18 recordsLinked to original sources

arXiv2026

From Diagnosis to Redesign: Using Quantitative Ethnography to Improve Multi-Agent LLM Reasoning

Multi-agent large language model (LLM) systems are designed to improve reasoning by decomposing tasks across multiple agents with specialized functions, but the presence of multiple agents does not inherently guarantee coherent reasoning or outputs that align with task objectives. This paper introduces a quantitative ethnographic (QE) approach for diagnosing and redesigning multi-agent LLM systems based on the discourse produced through agent interactions. We test this approach using automated essay scoring as an example context, applying Epistemic Network Analysis (ENA) to model a five-agent multi-agent debate system and examine differences between debates that produced correct versus incorrect scoring decisions. Results show that, in the initial system, correct scoring decisions were characterized by rubric-grounded justification, agreement, and elaboration. Incorrect scoring decisions, in contrast, were characterized by extended proposition-challenge-response exchanges that were less consistently tied to rubric criteria. We then used the findings to revise the agents' prompts. The revised system improved exact scoring accuracy from 27.78% to 40.28% and shifted the discourse of incorrect debates toward the rubric-grounded pattern of correct ones, making the two nearly indistinguishable. Based on these results, we argue that QE can support a diagnostic-to-redesign loop for AI reasoning by tracing how patterns of agent interaction relate to system performance, informing prompt redesign, and evaluating whether those redesigns change both outcomes and interaction patterns.

cs.CL↗

arXiv2026

Enhancing LLM-Based Data Annotation with Error Decomposition

Large language models offer a scalable alternative to human coding for data annotation tasks, enabling the scale-up of research across data-intensive domains. While LLMs are already achieving near-human accuracy on objective annotation tasks, their performance on subjective annotation tasks, such as those involving psychological constructs, is less consistent and more prone to errors. Standard evaluation practices typically collapse all annotation errors into a single alignment metric, but this simplified approach may obscure different kinds of errors that affect final analytical conclusions in different ways. Here, we propose a diagnostic evaluation paradigm that incorporates a human-in-the-loop step to separate task-inherent ambiguity from model-driven inaccuracies and assess annotation quality in terms of their potential downstream impacts. We refine this paradigm on ordinal annotation tasks, which are common in subjective annotation. The refined paradigm includes: (1) a diagnostic taxonomy that categorizes LLM annotation errors along two dimensions: source (model-specific vs. task-inherent) and type (boundary ambiguity vs. conceptual misidentification); (2) a lightweight human annotation test to estimate task-inherent ambiguity from LLM annotations; and (3) a computational method to decompose observed LLM annotation errors following our taxonomy. We validate this paradigm on four educational annotation tasks, demonstrating both its conceptual validity and practical utility. Theoretically, our work provides empirical evidence for why excessively high alignment is unrealistic in specific annotation tasks and why single alignment metrics inadequately reflect the quality of LLM annotations. In practice, our paradigm can be a low-cost diagnostic tool that assesses the suitability of a given task for LLM annotation and provides actionable insights for further technical optimization.

cs.CL↗

arXiv2026

Evaluating 21st-Century Competencies in Postsecondary Curricula with Large Language Models: Performance Benchmarking and Reasoning-Based Prompting Strategies

The growing emphasis on 21st-century competencies in postsecondary education, intensified by the transformative impact of generative AI, underscores the need to evaluate how these competencies are embedded in curricula and how effectively academic programs align with evolving workforce and societal demands. Curricular Analytics, particularly recent generative AI-powered approaches, offer a promising data-driven pathway. However, analyzing 21st-century competencies requires pedagogical reasoning beyond surface-level information retrieval, and the capabilities of large language models in this context remain underexplored. In this study, we extend prior curricular analytics research by examining a broader range of curriculum documents, competency frameworks, and models. Using 7,600 manually annotated curriculum-competency alignment scores, we assess the informativeness of different curriculum sources, benchmark general-purpose LLMs for curriculum-to-competency mapping, and analyze error patterns. We further introduce a reasoning-based prompting strategy, Curricular CoT, to strengthen LLMs' pedagogical reasoning. Our results show that detailed instructional activity descriptions are the most informative type of curriculum document for competency analytics. Open-weight LLMs achieve accuracy comparable to proprietary models on coarse-grained tasks, demonstrating their scalability and cost-effectiveness for institutional use. However, no model reaches human-level precision in fine-grained pedagogical reasoning. Our proposed Curricular CoT yields modest improvements by reducing bias in instructional keyword inference and improving the detection of nuanced pedagogical evidence in long text. Together, these findings highlight the untapped potential of institutional curriculum documents and provide an empirical foundation for advancing AI-driven curricular analytics.

cs.CY↗

arXiv2026

AI-exposed jobs deteriorated before ChatGPT

Public debate links worsening job prospects for AI-exposed occupations to the release of ChatGPT in late 2022. Using monthly U.S. unemployment insurance records, we measure occupation- and location-specific unemployment risk and find that risk rose in AI-exposed occupations beginning in early 2022, months before ChatGPT. Analyzing millions of LinkedIn profiles, we show that graduate cohorts from 2021 onward entered AI-exposed jobs at lower rates than earlier cohorts, with gaps opening before late 2022. Finally, from millions of university syllabi, we find that graduates taking more AI-exposed curricula had higher first-job pay and shorter job searches after ChatGPT. Together, these results point to forces pre-dating generative AI and to the ongoing value of LLM-relevant education.

econ.GN↗

arXiv2025

Bringing Pedagogy into Focus: Evaluating Virtual Teaching Assistants' Question-Answering in Asynchronous Learning Environments

Asynchronous learning environments (ALEs) are widely adopted for formal and informal learning, but timely and personalized support is often limited. In this context, Virtual Teaching Assistants (VTAs) can potentially reduce the workload of instructors, but rigorous and pedagogically sound evaluation is essential. Existing assessments often rely on surface-level metrics and lack sufficient grounding in educational theories, making it difficult to meaningfully compare the pedagogical effectiveness of different VTA systems. To bridge this gap, we propose an evaluation framework rooted in learning sciences and tailored to asynchronous forum discussions, a common VTA deployment context in ALE. We construct classifiers using expert annotations of VTA responses on a diverse set of forum posts. We evaluate the effectiveness of our classifiers, identifying approaches that improve accuracy as well as challenges that hinder generalization. Our work establishes a foundation for theory-driven evaluation of VTA systems, paving the way for more pedagogically effective AI in education.

cs.CL↗

arXiv2025

When the Past Misleads: Rethinking Training Data Expansion Under Temporal Distribution Shifts

Predictive models are typically trained on historical data to predict future outcomes. While it is commonly assumed that training on more historical data would improve model performance and robustness, data distribution shifts over time may undermine these benefits. This study examines how expanding historical data training windows under covariate shifts (changes in feature distributions) and concept shifts (changes in feature-outcome relationships) affects the performance and algorithmic fairness of predictive models. First, we perform a simulation study to explore scenarios with varying degrees of covariate and concept shifts in training data. Absent distribution shifts, we observe performance gains from longer training windows though they reach a plateau quickly; in the presence of concept shift, performance may actually decline. Covariate shifts alone do not significantly affect model performance, but may complicate the impact of concept shifts. In terms of fairness, models produce more biased predictions when the magnitude of concept shifts differs across sociodemographic groups; for intersectional groups, these effects are more complex and not simply additive. Second, we conduct an empirical case study of student retention prediction, a common machine learning application in education, using 12 years of student records from 23 minority-serving community colleges in the United States. We find concept shifts to be a key contributor to performance degradation when expanding the training window. Moreover, model fairness is compromised when marginalized populations have distinct data distribution shift patterns from their peers. Overall, our findings caution against conventional wisdom that "more data is better" and underscore the importance of using historical data judiciously, especially when it may be subject to data distribution shifts, to improve model performance and fairness.

cs.CY↗

arXiv2025

From Course to Skill: Evaluating LLM Performance in Curricular Analytics

Curricular analytics (CA) -- systematic analysis of curricula data to inform program and course refinement -- becomes an increasingly valuable tool to help institutions align academic offerings with evolving societal and economic demands. Large language models (LLMs) are promising for handling large-scale, unstructured curriculum data, but it remains uncertain how reliably LLMs can perform CA tasks. In this paper, we systematically evaluate four text alignment strategies based on LLMs or traditional NLP methods for skill extraction, a core task in CA. Using a stratified sample of 400 curriculum documents of different types and a human-LLM collaborative evaluation framework, we find that retrieval-augmented generation (RAG) is the top-performing strategy across all types of curriculum documents, while zero-shot prompting performs worse than traditional NLP methods in most cases. Our findings highlight the promise of LLMs in analyzing brief and abstract curriculum documents, but also reveal that their performance can vary significantly depending on model selection and prompting strategies. This underscores the importance of carefully evaluating the performance of LLM-based strategies before large-scale deployment.

cs.CY↗

arXiv2025

Towards Fair and Privacy-Aware Transfer Learning for Educational Predictive Modeling: A Case Study on Retention Prediction in Community Colleges

Predictive analytics is widely used in learning analytics, but many resource-constrained institutions lack the capacity to develop their own models or rely on proprietary ones trained in different contexts with little transparency. Transfer learning holds promise for expanding equitable access to predictive analytics but remains underexplored due to legal and technical constraints. This paper examines transfer learning strategies for retention prediction at U.S. two-year community colleges. We envision a scenario where community colleges collaborate with each other and four-year universities to develop retention prediction models under privacy constraints and evaluate risks and improvement strategies of cross-institutional model transfer. Using administrative records from 4 research universities and 23 community colleges covering over 800,000 students across 7 cohorts, we identify performance and fairness degradation when external models are deployed locally without adaptation. Publicly available contextual information can forecast these performance drops and offer early guidance for model portability. For developers under privacy regulations, sequential training selecting institutions based on demographic similarities enhances fairness without compromising performance. For institutions lacking local data to fine-tune source models, customizing evaluation thresholds for sensitive groups outperforms standard transfer techniques in improving performance and fairness. Our findings suggest the value of transfer learning for more accessible educational predictive modeling and call for judicious use of contextual information in model training, selection, and deployment to achieve reliable and equitable model transfer.

cs.CY↗

arXiv2024

Whose ChatGPT? Unveiling Real-World Educational Inequalities Introduced by Large Language Models

The universal availability of ChatGPT and other similar tools since late 2022 has prompted tremendous public excitement and experimental effort about the potential of large language models (LLMs) to improve learning experience and outcomes, especially for learners from disadvantaged backgrounds. However, little research has systematically examined the real-world impacts of LLM availability on educational equity beyond theoretical projections and controlled studies of innovative LLM applications. To depict trends of post-LLM inequalities, we analyze 1,140,328 academic writing submissions from 16,791 college students across 2,391 courses between 2021 and 2024 at a public, minority-serving institution in the US. We find that students' overall writing quality gradually increased following the availability of LLMs and that the writing quality gaps between linguistically advantaged and disadvantaged students became increasingly narrower. However, this equitizing effect was more concentrated on students with higher socioeconomic status. These findings shed light on the digital divides in the era of LLMs and raise questions about the equity benefits of LLMs in early stages and highlight the need for researchers and practitioners on developing responsible practices to improve educational equity through LLMs.

cs.CY↗

arXiv2024

The Life Cycle of Large Language Models: A Review of Biases in Education

Large Language Models (LLMs) are increasingly adopted in educational contexts to provide personalized support to students and teachers. The unprecedented capacity of LLM-based applications to understand and generate natural language can potentially improve instructional effectiveness and learning outcomes, but the integration of LLMs in education technology has renewed concerns over algorithmic bias which may exacerbate educational inequities. In this review, building on prior work on mapping the traditional machine learning life cycle, we provide a holistic map of the LLM life cycle from the initial development of LLMs to customizing pre-trained models for various applications in educational settings. We explain each step in the LLM life cycle and identify potential sources of bias that may arise in the context of education. We discuss why current measures of bias from traditional machine learning fail to transfer to LLM-generated content in education, such as tutoring conversations because the text is high-dimensional, there can be multiple correct responses, and tailoring responses may be pedagogically desirable rather than unfair. This review aims to clarify the complex nature of bias in LLM applications and provide practical guidance for their evaluation to promote educational equity.

cs.CY↗

arXiv2024

Fairness Hub Technical Briefs: Definition and Detection of Distribution Shift

Distribution shift is a common situation in machine learning tasks, where the data used for training a model is different from the data the model is applied to in the real world. This issue arises across multiple technical settings: from standard prediction tasks, to time-series forecasting, and to more recent applications of large language models (LLMs). This mismatch can lead to performance reductions, and can be related to a multiplicity of factors: sampling issues and non-representative data, changes in the environment or policies, or the emergence of previously unseen scenarios. This brief focuses on the definition and detection of distribution shifts in educational settings. We focus on standard prediction problems, where the task is to learn a model that takes in a series of input (predictors) $X=(x_1,x_2,...,x_m)$ and produces an output $Y=f(X)$.

cs.LG↗

arXiv2024

Course-Skill Atlas: A national longitudinal dataset of skills taught in U.S. higher education curricula

Higher education plays a critical role in driving an innovative economy by equipping students with knowledge and skills demanded by the workforce. While researchers and practitioners have developed data systems to track detailed occupational skills, such as those established by the U.S. Department of Labor (DOL), much less effort has been made to document which of these skills are being developed in higher education at a similar granularity. Here, we fill this gap by presenting Course-Skill Atlas -- a longitudinal dataset of skills inferred from over three million course syllabi taught at nearly three thousand U.S. higher education institutions. To construct Course-Skill Atlas, we apply natural language processing to quantify the alignment between course syllabi and detailed workplace activities (DWAs) used by the DOL to describe occupations. We then aggregate these alignment scores to create skill profiles for institutions and academic majors. Our dataset offers a large-scale representation of college education's role in preparing students for the labor market. Overall, Course-Skill Atlas can enable new research on the source of skills in the context of workforce development and provide actionable insights for shaping the future of higher education to meet evolving labor demands, especially in the face of new technologies.

econ.GN↗

arXiv2024

Temporal and Between-Group Variability in College Dropout Prediction

Large-scale administrative data is a common input in early warning systems for college dropout in higher education. Still, the terminology and methodology vary significantly across existing studies, and the implications of different modeling decisions are not fully understood. This study provides a systematic evaluation of contributing factors and predictive performance of machine learning models over time and across different student groups. Drawing on twelve years of administrative data at a large public university in the US, we find that dropout prediction at the end of the second year has a 20% higher AUC than at the time of enrollment in a Random Forest model. Also, most predictive factors at the time of enrollment, including demographics and high school performance, are quickly superseded in predictive importance by college performance and in later stages by enrollment behavior. Regarding variability across student groups, college GPA has more predictive value for students from traditionally disadvantaged backgrounds than their peers. These results can help researchers and administrators understand the comparative value of different data sources when building early warning systems and optimizing decisions under specific policy goals.

cs.CY↗

arXiv2023

Fairness Hub Technical Briefs: AUC Gap

To measure bias, we encourage teams to consider using AUC Gap: the absolute difference between the highest and lowest test AUC for subgroups (e.g., gender, race, SES, prior knowledge). It is agnostic to the AI/ML algorithm used and it captures the disparity in model performance for any number of subgroups, which enables non-binary fairness assessments such as for intersectional identity groups. The teams use a wide range of AI/ML models in pursuit of a common goal of doubling math achievement in low-income middle schools. Ensuring that the models, which are trained on datasets collected in many different contexts, do not introduce or amplify biases is important for achieving the goal. We offer here a versatile and easy-to-compute measure of model bias for all the teams in order to create a common benchmark and an analytical basis for sharing what strategies have worked for different teams.

cs.LG↗

arXiv2023

Cross-Institutional Transfer Learning for Educational Models: Implications for Model Performance, Fairness, and Equity

Modern machine learning increasingly supports paradigms that are multi-institutional (using data from multiple institutions during training) or cross-institutional (using models from multiple institutions for inference), but the empirical effects of these paradigms are not well understood. This study investigates cross-institutional learning via an empirical case study in higher education. We propose a framework and metrics for assessing the utility and fairness of student dropout prediction models that are transferred across institutions. We examine the feasibility of cross-institutional transfer under real-world data- and model-sharing constraints, quantifying model biases for intersectional student identities, characterizing potential disparate impact due to these biases, and investigating the impact of various cross-institutional ensembling approaches on fairness and overall model performance. We perform this analysis on data representing over 200,000 enrolled students annually from four universities without sharing training data between institutions. We find that a simple zero-shot cross-institutional transfer procedure can achieve similar performance to locally-trained models for all institutions in our study, without sacrificing model fairness. We also find that stacked ensembling provides no additional benefits to overall performance or fairness compared to either a local model or the zero-shot transfer procedure we tested. We find no evidence of a fairness-accuracy tradeoff across dozens of models and transfer schemes evaluated. Our auditing procedure also highlights the importance of intersectional fairness analysis, revealing performance disparities at the intersection of sensitive identity groups that are concealed under one-dimensional analysis.

cs.LG↗

arXiv2022

A Robust Approach for the Decomposition of High-Energy-Consuming Industrial Loads with Deep Learning

The knowledge of the users' electricity consumption pattern is an important coordinating mechanism between the utility company and the electricity consumers in terms of key decision makings. The load decomposition is therefore crucial to reveal the underlying relationship between the load consumption and its characteristics. However, load decomposition is conventionally performed on the residential and commercial loads, and adequate consideration has not been given to the high-energy-consuming industrial loads leading to inefficient results. This paper thus focuses on the load decomposition of the industrial park loads (IPL). The commonly used parameters in a conventional method are however inapplicable in high-energy-consuming industrial loads. Therefore, a more robust approach is developed comprising a three-algorithm model to achieve this goal on the IPL. First, the improved variational mode decomposition (IVMD) algorithm is introduced to denoise the training data of the IPL and improve its stability. Secondly, the convolutional neural network (CNN) and simple recurrent units (SRU) joint algorithms are used to achieve a non-intrusive and non-invasive decomposition process of the IPL using a double-layer deep learning network based on the IPL characteristics. Specifically, CNN is used to extract the IPL data characteristics while the improved long and short-term memory (LSTM) network, SRU, is adopted to develop the decomposition model and further train the load data. Through the robust decomposition process, the underlying relationship in the load consumption is extracted. The results obtained from the numerical examples show that this approach outperforms the state-of-the-art in the conventional decomposition process.

cs.LG↗

arXiv2021

Should College Dropout Prediction Models Include Protected Attributes?

Early identification of college dropouts can provide tremendous value for improving student success and institutional effectiveness, and predictive analytics are increasingly used for this purpose. However, ethical concerns have emerged about whether including protected attributes in the prediction models discriminates against underrepresented student groups and exacerbates existing inequities. We examine this issue in the context of a large U.S. research university with both residential and fully online degree-seeking students. Based on comprehensive institutional records for this entire student population across multiple years, we build machine learning models to predict student dropout after one academic year of study, and compare the overall performance and fairness of model predictions with or without four protected attributes (gender, URM, first-generation student, and high financial need). We find that including protected attributes does not impact the overall prediction performance and it only marginally improves algorithmic fairness of predictions. While these findings suggest that including protected attributes is preferred, our analysis also offers guidance on how to evaluate the impact in a local context, where institutional stakeholders seek to leverage predictive analytics to support student success.

cs.CY↗

arXiv2021

Unsupervised Representations Predict Popularity of Peer-Shared Artifacts in an Online Learning Environment

In online collaborative learning environments, students create content and construct their own knowledge through complex interactions over time. To facilitate effective social learning and inclusive participation in this context, insights are needed into the correspondence between student-contributed artifacts and their subsequent popularity among peers. In this study, we represent student artifacts by their (a) contextual action logs (b) textual content, and (c) set of instructor-specified features, and use these representations to predict artifact popularity measures. Through a mixture of predictive analysis and visual exploration, we find that the neural embedding representation, learned from contextual action logs, has the strongest predictions of popularity, ahead of instructor's knowledge, which includes academic value and creativity ratings. Because this representation can be learnt without extensive human labeling effort, it opens up possibilities for shaping more inclusive student interactions on the fly in collaboration with instructors and students alike.

cs.CY↗