Hanchen David Wang

Clayton Cohn, Surya Rayala, Siyuan Guo et al.

LLM pedagogical agents are proliferating, yet recent findings have raised questions about their adherence to established theories of learning and, by extension, their educational value. Concerns regarding cognitive offloading, over-reliance, and "gaming" behaviors persist and remain largely unaddressed. In response, we developed Copa, an agentic, multi-agent, multimodal Collaborative Peer Agent for STEM+C learning. Copa is built on top of the Evidence-Decision-Feedback (EDF) framework, grounding its interactions in Social Cognitive Theory and Social Constructivism and promoting sense-making through adaptive, dialogic support rather than answer-seeking. In an authentic high school computational-modeling study (n=33 dyads), we demonstrate that Copa (1) supports students' confidence building and ability to verbalize conceptual understanding without causing dependence; and (2) provides adaptive feedback personalized to learners that is interpretable with respect to students' multimodal input data. These findings position theory-guided, multimodal LLM agents as a promising path toward classroom AI integration that amplifies students' reasoning rather than replacing it.

Hanchen David Wang, Diego Manzanas Lopez, Preston K. Robinette et al.

As deep neural networks are deployed in safety-critical domains such as autonomous driving and medical diagnosis, stakeholders need explanations that are interpretable but also trustworthy with formal guarantees. Existing XAI methods fall short: heuristic attribution techniques (e.g., LIME, Integrated Gradients) highlight influential features but offer no mathematical guarantees about decision boundaries, while formal methods verify robustness yet remain untargeted, analyzing the nearest boundary regardless of whether it represents a critical risk. In safety-critical systems, not all misclassifications carry equal consequences; confusing a "Stop" sign for a "60 kph" sign is far more dangerous than confusing it with a "No Passing" sign. We introduce ViTaX (Verified and Targeted Explanations), a formal XAI framework that generates targeted semifactual explanations with mathematical guarantees. For a given input (class y) and a user-specified critical alternative (class t), ViTaX: (1) identifies the minimal feature subset most sensitive to the y->t transition, and (2) applies formal reachability analysis to guarantee that perturbing these features by epsilon cannot flip the classification to t. We formalize this through Targeted epsilon-Robustness, certifying whether a feature subset remains robust under perturbation toward a specific target class. ViTaX is the first method to provide formally guaranteed explanations of a model's resilience against user-identified alternatives. Evaluations on MNIST, GTSRB, EMNIST, and TaxiNet demonstrate over 30% fidelity improvement with minimal explanation cardinality.

Hanchen David Wang, Meiyi Ma

Physical therapy (PT) is crucial for patients to restore and maintain mobility, function, and well-being. Many on-site activities and body exercises are performed under the supervision of therapists or clinicians. However, the postures of some exercises at home cannot be performed accurately due to the lack of supervision, quality assessment, and self-correction. Therefore, in this paper, we design a new framework, PhysiQ, that continuously tracks and quantitatively measures people's off-site exercise activity through passive sensory detection. In the framework, we create a novel multi-task spatio-temporal Siamese Neural Network that measures the absolute quality through classification and relative quality based on an individual's PT progress through similarity comparison. PhysiQ digitizes and evaluates exercises in three different metrics: range of motions, stability, and repetition.

Clayton Cohn, Eduardo Davalos, Caleb Vatral et al.

Recent technological advancements in multimodal machine learning--including the rise of large language models (LLMs)--have improved our ability to collect, process, and analyze diverse multimodal data such as speech, video, and eye gaze in learning and training contexts. While prior reviews have addressed individual components of the multimodal pipeline (e.g., conceptual models, data fusion), a comprehensive review of empirical methods in applied multimodal environments remains notably absent. This review addresses that, introducing a taxonomy and framework that capture both established practices and recent innovations driven by LLMs and generative AI. We identify five modality groups: Natural Language, Vision, Physiological Signals, Human-Centered Evidence, and Environment Logs. Our analysis reveals that integrating modalities enables richer insights into learner and trainee behaviors, revealing latent patterns often overlooked by unimodal approaches. However, persistent challenges in multimodal data collection and integration continue to hinder the adoption of these systems in real-time classroom settings.

Clayton Cohn, Siyuan Guo, Surya Rayala et al.

Multi-agent LLM architectures offer opportunities for pedagogical agents to help students construct domain knowledge and develop critical-thinking skills, yet many operate on a "one-size-fits-all" basis, limiting their ability to provide personalized support. To address this, we introduce Evidence-Decision-Feedback (EDF), a theoretical framework for adaptive scaffolding using LLMs. EDF integrates elements of intelligent tutoring systems and agentic behavior by organizing interactions around evidentiary inference, pedagogical decision-making, and adaptive feedback. We instantiate EDF through Copa, an agentic collaborative peer agent for STEM+C problem-solving. In an authentic high school classroom study, we show that EDF-guided interactions align feedback with students' demonstrated understanding and task mastery; promote gradual scaffold fading; and support interpretable, evidence-grounded explanations without fostering overreliance.

Hanchen David Wang, Clayton Cohn, Zifan Xu et al.

Simulating student learning behaviors in open-ended problem-solving environments holds potential for education research, from training adaptive tutoring systems to stress-testing pedagogical interventions. However, collecting authentic data is challenging due to privacy concerns and the high cost of longitudinal studies. While Large Language Models (LLMs) offer a promising path to student simulation, they suffer from competency bias, optimizing for efficient correctness rather than the erratic, iterative struggle characteristic of novice learners. We present BEAGLE, a neuro-symbolic framework that addresses this bias by incorporating Self-Regulated Learning (SRL) theory into a novel architecture. BEAGLE integrates three key technical innovations: (1) a semi-Markov model that governs the timing and transitions of cognitive behaviors and metacognitive behaviors; (2) Bayesian Knowledge Tracing with explicit flaw injection to enforce realistic knowledge gaps and "unknown unknowns"; and (3) a decoupled agent design that separates high-level strategy use from code generation actions to prevent the model from silently correcting its own intentional errors. In evaluations on Python programming tasks, BEAGLE significantly outperforms state-of-the-art baselines in reproducing authentic trajectories. In a human Turing test, users were unable to distinguish synthetic traces from real student data, achieving an accuracy indistinguishable from random guessing (52.8%).

Hanchen David Wang, Yilin Liu, Madison J. Lee et al.

Assessing learner competency in clinical simulation requires expert observation that is time-intensive, difficult to scale, and subject to inter-rater variability. Vision-language models have emerged as a promising tool for understanding complex visual behavior. In this work, we investigate whether visual observations can provide educationally meaningful signals for competency assessment through a three-stage framework that (1) extracts action timelines from egocentric nursing simulation video using frozen visual encoders and few-shot learning, (2) derives sequence-level features and per-session recognition metrics, and (3) relates these to instructor-rated competency. Across 22 densely annotated sessions (3.8 hours, 493 actions), a frozen DINOv2 backbone with HMM Viterbi decoding achieves 57.4% MOF in leave-one-out 1-shot recognition. Surprisingly, we observe a negative trend between recognition accuracy and competency (rho = -0.524, p = 0.012 for mIoU), robust to six confound controls: more competent students produce diverse, harder-to-classify workflows, while simple sequence features show no such relationship. Per-item analysis identifies patient safety protocols and team communication as the expected behaviors most reflected in this pattern, and process model comparisons reveal that higher-competency students exhibit more protocol-consistent action transitions. These findings suggest that recognition accuracy may complement predicted action timelines as a pedagogically informative signal in automated competency assessment.

Hanchen David Wang, Nibraas Khan, Anna Chen et al.

Recent global estimates suggest that as many as 2.41 billion individuals have health conditions that would benefit from rehabilitation services. Home-based Physical Therapy (PT) faces significant challenges in providing interactive feedback and meaningful observation for therapists and patients. To fill this gap, we present MicroXercise, which integrates micro-motion analysis with wearable sensors, providing therapists and patients with a comprehensive feedback interface, including video, text, and scores. Crucially, it employs multi-dimensional Dynamic Time Warping (DTW) and attribution-based explainable methods to analyze the existing deep learning neural networks in monitoring exercises, focusing on a high granularity of exercise. This synergistic approach is pivotal, providing output matching the input size to precisely highlight critical subtleties and movements in PT, thus transforming complex AI analysis into clear, actionable feedback. By highlighting these micro-motions in different metrics, such as stability and range of motion, MicroXercise significantly enhances the understanding and relevance of feedback for end-users. Comparative performance metrics underscore its effectiveness over traditional methods, such as a 39% and 42% improvement in Feature Mutual Information (FMI) and Continuity. MicroXercise is a step ahead in home-based physical therapy, providing a technologically advanced and intuitively helpful solution to enhance patient care and outcomes.

Hanchen David Wang, Siwoo Bae, Zirong Chen et al.

Artificial intelligence systems in critical fields like autonomous driving and medical imaging analysis often continually learn new tasks using a shared stream of input data. For instance, after learning to detect traffic signs, a model may later need to learn to classify traffic lights or different types of vehicles using the same camera feed. This scenario introduces a challenging setting we term Continual Multitask Learning (CMTL), where a model sequentially learns new tasks on an underlying data distribution without forgetting previously learned abilities. Existing continual learning methods often fail in this setting because they learn fragmented, task-specific features that interfere with one another. To address this, we introduce Learning with Preserving (LwP), a novel framework that shifts the focus from preserving task outputs to maintaining the geometric structure of the shared representation space. The core of LwP is a Dynamically Weighted Distance Preservation (DWDP) loss that prevents representation drift by regularizing the pairwise distances between latent data representations. This mechanism of preserving the underlying geometric structure allows the model to retain implicit knowledge and support diverse tasks without requiring a replay buffer, making it suitable for privacy-conscious applications. Extensive evaluations on time-series and image benchmarks show that LwP not only mitigates catastrophic forgetting but also consistently outperforms state-of-the-art baselines in CMTL tasks. Notably, our method shows superior robustness to distribution shifts and is the only approach to surpass the strong single-task learning baseline, underscoring its effectiveness for real-world dynamic environments.