arXiv subjects

Shayla Sharmin

Publications and source records attributed to Shayla Sharmin.

17 recordsLinked to original sources

arXiv2026

Do Video-LLMs Actually Watch? Diagnosing Character-Tracking Failures in Long-Form Video

Can a Video Large Language Model (Video-LLM) follow one person through a long video, keeping track of who they are well enough to report, in order, how their outfit changes across a full TV episode? Benchmarks increasingly score this kind of task, and the strongest open-source 7--8B models now reach 37--38% on InfiniBench's global appearance task, which asks exactly that. But does that score come from tracking the named character, or from something easier? We test this with a nine-condition diagnostic protocol applied to three architecturally distinct open-source Video-LLMs, with Gemini~2.5~Flash as a frontier reference, and find the accuracy does not come from character tracking. When we change the character named in the question to a different cast member, leaving the video and answer options untouched, the models change their answer only 4--31% of the time, so they are largely ignoring who the question asks about. Breaking that test down by the gender of the swapped name shows why: the models react more when the name is changed to a different-gender character than to a same-gender one (a 13--28 point gap), picking up coarse gender cues but unable to tell same-gender individuals apart. This shallow processing surfaces again when we drop the multiple-choice options and ask the same questions open-endedly: open-source accuracy drops 18--25 points, with none of 151 answers fully correct, versus a 12-point drop for Gemini. Further checks rule out the obvious innocent explanations, adding subtitles, using the most informative frames, or doubling the number of frames all leave character tracking unimproved, so the bottleneck is not how much video the model sees but how it ties that video to the person the question names. We release a diagnostic toolkit for auditing what such benchmark scores actually measure.

cs.CV↗

arXiv2026

Indirect and Direct AI Scaffolding for Computational Problem Posing: A Pilot Experience Report

Problem posing is a valuable learning activity in computing education, encouraging learners to actively construct, refine, and reflect on problems rather than simply solving them. This experience report presents the design and pilot deployment of two LLM-powered scaffolding systems for supporting problem posing across two computational scenarios with different levels of task openness. Both systems assessed student-generated problems using Bloom's Taxonomy-based criteria and applied the same assessment framework, differing only in output modality: one provided guiding questions (Indirect scaffolding), while the other offered worked examples (Direct scaffolding). We conducted a within-subjects, counterbalanced pilot study with 20 graduate students and collected problem-quality ratings, user-experience surveys, and post-session interviews. Our deployment showed that both systems supported problem refinement in complementary ways, each offering distinct benefits. Direct scaffolding produced greater immediate improvements, while interviews showed that participants valued Indirect scaffolding for promoting deeper reflection on their own problem design. Based on these findings, we suggest sequencing the two modalities by beginning with Indirect scaffolding to promote reflection, then shifting to Direct scaffolding when learners become stuck. These lessons offer an initial practical strategy for integrating LLM-based scaffolding into computing classrooms.

cs.HC↗

arXiv2026

When LLM Tutoring Responses Work: Evidence from Student Programming Conversations

As students increasingly use LLM tutors in computer science education, one question becomes especially important: what kind of response helps a student continue productively? Prior work has studied how students use LLMs in computer science education, but less is known about how tutoring response styles are associated with student follow-up across programming help-seeking contexts. This paper analyzes StudyChat (UMass, 2026), a public dataset of student and ChatGPT tutoring conversations from an artificial intelligence course. We transformed StudyChat into 16,851 assistant-response interactions from 203 students and 2,214 conversations. Using local LLM-assisted annotation with Gemma 4, we labeled student help-seeking situations, student state, assistant response style, and student next-turn outcome. Human validation showed 82\% agreement with the LLM-assisted labels (Cohen's $\kappa=.74$). We analyzed productive continuation and unresolved continuation across the full dataset and across help-seeking contexts. Globally, response style was significantly associated with productive continuation, $\chi^2(7)=100.39$, $p<.001$, $V=.078$, and unresolved continuation, $\chi^2(7)=125.77$, $p<.001$, $V=.087$, though effect sizes were small. Verification feedback had the highest productive-continuation rate (82.4\%), while direct answers had the lowest (62.7\%). Descriptively, response-style score ranges were smallest in low-confusion conceptual contexts (.017) and largest in high-cognitive-load contexts (.203). More detailed comparisons showed situation-dependent response patterns. For example, stepwise guidance was followed by greater confusion decrease in high-cognitive-load code requests, while direct answers were followed by more unresolved continuation in high-load debugging. These findings support context-aware evaluation and design of AI tutoring responses for programming education.

cs.HC↗

arXiv2026

Reading the Eyes in VR: Multimodal Modeling of Social Intelligence

Social intelligence, the ability to interpret others' emotions, beliefs, and intentions, is often assessed with the Reading the Mind in the Eyes Test (RMET), in which participants infer mental states from images of the eye region. Yet RMET is typically presented on paper or desktop displays, where viewing geometry can vary across participants, and it rarely includes immediate feedback. We investigated whether presentation medium and brief trial-level feedback influence RMET behavior. We implemented RMET in Unity for both desktop and Virtual Reality (VR), using VR to hold stimulus distance and field of view constant without changing the items. We conducted a 2x2 mixed study with 20 participants, with device (VR vs. desktop) manipulated between subjects and feedback (immediate correctness cue vs. none) manipulated within subjects. Eye-tracking and EEG data were recorded and synchronized with behavioral logs. We analyzed fixation-based gaze measures, EEG signals, response time, accuracy, and subjective measures. Immediate feedback was associated with longer fixation durations and higher EEG-based engagement, while no significant differences were observed in task completion time or total correct answers. Presentation medium did not produce reliable differences in the objective measures, but VR received higher usability ratings and was also rated as more effortful. These results provide initial evidence that RMET can be studied as a process-aware assessment task in controlled VR and desktop settings.

cs.HC↗

arXiv2026

How YouTube Frames ChatGPT Use in Education: An Epistemic Network Analysis with Supporting Multimodal Metadata

We examine educational YouTube videos through multimodal metadata, such as transcripts, titles, thumbnails, and viewer comments, to investigate how ChatGPT is framed across creator groups and how those framings relate to audience response and platform reach. Little is known about how large language models are presented to learners in informal, creator-driven public discourse. Following PRISMA, we selected 52 videos for analysis. We identified three structurally distinct discourse groups: (G1) videos that positioned ChatGPT as a conceptual scaffold for thinking, (G2) videos oriented toward retrieval practice and skill-building, and (G3) videos that framed ChatGPT as a tool for output generation. Epistemic Network Analysis revealed statistically significant group differences with large effect sizes. Multimodal metadata consistently reflected these distinctions across transcript discourse, titles, and thumbnails. Viewers of learning-oriented content described ChatGPT as a thinking partner or tutor, whereas viewers of output-oriented content raised concerns about over-reliance, surface-level learning, and cognitive offloading. G3 achieved comparable platform reach to G2, yet with substantially weaker learning-oriented framing. This may suggest that output-oriented content competes for visibility despite lower pedagogical depth. These findings reveal a structural tension in self-directed AI learning: content that prioritizes quick outputs reaches far more learners than content that promotes deep engagement. This gap raises critical questions about whose vision of AI literacy scales and what learners are actually left with.

cs.HC↗

arXiv2026

Micro-level AI Feedback Features and Student Responses in Consecutive LLM Tutoring Interactions

AI-assisted feedback research has shown that micro-level feedback features, such as concrete elaboration, affective language, and response length, are associated with learning outcomes. Existing studies have primarily examined these features using session- or task-level measures. We examine how feedback provided in one user-AI interaction is associated with student confusion and understanding in the immediately following interaction in a naturalistic tutoring setting. We focus on three micro-level features of AI feedback: concrete elaboration (analogies, comparison-based explanations, or worked examples), affective language (encouragement, empathy, or apology), and response length. We analyzed 16,851 conversational user-AI interactions from the StudyChat dataset, a naturalistic record of student interactions with an LLM tutor in an undergraduate AI course, and identified 1,718 cases in which students expressed confusion and continued to a subsequent interaction. Using chi-square tests and Generalized Estimating Equations (GEE), we found that concrete elaboration was associated with higher understanding and lower re-confusion in the student's next interaction. Empathetic language showed no significant association with either outcome, while longer responses were independently associated with lower understanding. These findings highlight the value of examining feedback across consecutive user-AI interactions and suggest that concrete elaboration may play an important role in supporting immediate student understanding.

cs.HC↗

arXiv2026

AFA: Identity-Aware Memory for Preventing Persona Confusion in Multi-User Dialogue

When multiple people share a single voice assistant, the system conflates their histories: one resident's preferences can leak into another's responses, eroding utility and trust. We call this failure mode persona confusion, and we show it is a measurable problem in today's single-user dialogue systems when deployed in shared environments. We present the Adaptive Friend Agent (AFA), a modular framework that combines voice-based speaker identification with per-user memory stores to enable identity-aware, personalized dialogue across multiple users. To support training and evaluation, we construct PAT (Personalized Agent chaT), a synthetic dataset of 58,289 persona-grounded dialogue turns spanning 133 user profiles and 12 real-world scenarios. We evaluate AFA across five LLM back-ends in a standard response-quality benchmark, with a LLaMA-2-70B model fine-tuned on PAT achieving the highest overall performance. To directly measure persona confusion prevention, we introduce an interleaved multi-user evaluation protocol with a novel metric, Persona Attribution Accuracy (PAA), demonstrating that identity-aware routing improves PAA from 35.7% to 61.3%. Human evaluation confirms annotators perceive significantly higher personalization in routing-enabled responses. Our results establish that identity-aware user routing is the critical component for preventing persona confusion in multi-user conversational systems.

cs.HC↗

arXiv2026

Avoiding Social Judgment, Seeking Privacy: Investigating why Mothers Shift from Facebook Groups to Large Language Models

Social media platforms, especially Facebook parenting groups, have long been used as informal support networks for mothers seeking advice and reassurance. However, growing concerns about social judgment, privacy exposure, and unreliable information are changing how mothers seek help. This exploratory mixed-method study examines why mothers are moving from Facebook parenting groups to large language models such as ChatGPT and Gemini. We conducted a cross-sectional online survey of 109 mothers. Results show that 41.3% of participants avoided Facebook parenting groups because they expected judgment from others. This difference was statistically significant across location and family structure. Mothers living in their home country and those in joint families were more likely to avoid Facebook groups. Qualitative findings revealed three themes: social judgment and exposure, LLMs as safe and private spaces, and quick and structured support. Participants described LLMs as immediate, emotionally safe, and reliable alternatives that reduce social risk when asking for help. Rather than replacing human support, LLMs appear to fill emotional and practical gaps within existing support systems. These findings show a change in maternal digital support and highlight the need to design LLM systems that support both information and emotional safety.

cs.HC↗

arXiv2026

Beyond Judgment: Exploring Large Language Models as Non-Judgmental Support for Maternal Mental Health

In the age of Large Language Models (LLMs), much work has already been done on how LLMs support medication advice and serve as information providers; however, how mothers use these tools for emotional and informational support to avoid social judgment remains underexplored. This study conducted a 10-day mixed-methods exploratory survey ($N=107$) to investigate how mothers use LLMs as a non-judgmental resource for emotional support and regulation, and for situational reassurance. Our findings show that mothers are asking LLMs various questions about childcare to reassure themselves and avoid judgment, particularly around childcare decisions, maternal guilt, and late-night caregiving. Open-ended responses also show that mothers are comfortable with LLMs because they do not have to think about social consequences or judgment. Although mothers use LLMs for quick information or reassurance to avoid judgment, over half of the participants value human warmth more than LLMs; however, a significant minority, especially those in joint families, consider LLMs to avoid human judgment. These findings help understand how LLMs can be framed as low-risk interaction support rather than a replacement for human support, and highlight the role of social context in shaping emotional technology use.

cs.HC↗

arXiv2025

Beyond Cognitive Load: AI-Based Estimation of Cognitive Effort Using Brain Signals During Digital Tasks

Cognitive effort, defined as the relationship between cognitive load and task performance, provides insight into how individuals allocate mental resources during demanding tasks. This construct is particularly important in high-stakes public health and clinical training, where excessive cognitive load is associated with medical errors and burnout. This study investigates whether cognitive effort varies across task segments and whether it can be estimated at the individual level using brain signal data and machine learning. Functional near-infrared spectroscopy (fNIRS) data were collected from 16 participants performing a structured digital cognitive task consisting of four sequential segments separated by short and long rest intervals. Cognitive effort was operationalized using relative neural efficiency and relative neural involvement, integrating prefrontal hemodynamic activity with task performance. The analysis followed a two-stage approach. First, segment-level group analysis tested whether cognitive effort differed across task segments, assessing whether the task structure induced meaningful variation in cognitive demand. Second, participant-independent machine learning models were used to predict task performance from brain signal features. These predicted scores were then combined with neural measures to estimate individual-level cognitive effort. Results showed significant differences in cognitive effort across the four task segments, indicating that variations in task structure influence collective cognitive efficiency. In addition, machine learning models successfully predicted performance from fNIRS data. Cognitive effort derived from predicted scores closely matched that based on actual performance, suggesting that the proposed metric primarily reflects brain signal patterns.

cs.HC↗

arXiv2025

Hybrid Deep Learning Model to Estimate Cognitive Effort from fNIRS Signals

This study estimates cognitive effort based on functional near-infrared spectroscopy data and performance scores using a hybrid DeepNet model. The estimation of cognitive effort enables educators to modify material to enhance learning effectiveness and student engagement. In this study, we collected oxygenated hemoglobin using functional near-infrared spectroscopy during an educational quiz game. Participants (n=16) responded to 16 questions in a Unity-based educational game, each within a 30-second response time limit. We used DeepNet models to predict the performance score from the oxygenated hemoglobin, and compared traditional machine learning and DeepNet models to determine which approach provides better accuracy in predicting performance scores. The result shows that the proposed CNN-GRU gives better performance with 73% than other models. After the prediction, we used the predicted score and the oxygenated hemoglobin to observe cognitive effort by calculating relative neural efficiency and involvement in our test cases. Our result shows that even with moderate accuracy, the predicted cognitive effort closely follow the actual trends. This findings can be helpful in designing and improving learning environments and provide valuable insights into learning materials.

cs.HC↗

arXiv2024

From Complexity to Simplicity: Using Python Instead of PsychoPy for fNIRS Data Collection

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical technique that measures brain activity by estimating blood oxygenation using near-infrared light. Traditionally, PsychoPy is used in many studies to send task-specific markers, requiring a separate device to interface with the fNIRS data collection system. In this work, we present a Python-based implementation to send markers directly, eliminating the need for an additional device. This approach allows researchers to run both marker transmission and fNIRS data collection on the same computer, simplifying the setup and enhancing accessibility. This streamlined solution reduces hardware requirements and makes fNIRS studies more efficient.

cs.HC↗

arXiv2024

Neurophysiological Insights into Multimedia-based Education: A PRISMA-ScR Review of fNIRS in Game-Integrated Learning Systems

Game-integrated learning systems (GILS) are a growing form of multimedia education. Brain-based evidence can help researchers and designers understand how GILS design choices shape how learners think and process information. This scoping review follows PRISMA-ScR and synthesizes 20 empirical studies (2014-2025) in which functional near-infrared spectroscopy (fNIRS) measured brain activity during GILS use. This corpus shows that fNIRS can capture brain responses across GILS platforms and game elements, and points to how neurophysiological evidence can inform multimedia design decisions, such as that different platforms activate different brain regions, that adaptive difficulty reduces cognitive load and improves performance simultaneously, and that collaborative gameplay predicts knowledge retention. The 20 studies in this corpus reflect a field with substantial room to grow. Causal links between brain activation and learning outcomes would give designers more reliable evidence for platform decisions. As fNIRS and multimedia devices improve, standardized methods, classroom settings, and real-time neural adaptation represent directions where future work can translate these findings into practical multimedia learning systems.

cs.HC↗

arXiv2024

Analyzing Nursing Assistant Attitudes Towards Empathic Geriatric Caregiving Using Quantitative Ethnography

An emergent challenge in geriatric care is improving the quality of care, which requires insight from stakeholders. Qualitative methods offer detailed insights, but they can be biased and have limited generalizability, while quantitative methods may miss nuances. Network-based approaches, such as quantitative ethnography (QE), can bridge this methodological gap. By leveraging the strengths of both methods, QE provides profound insights into need-finding interviews. In this paper, to better understand geriatric care attitudes, we interviewed ten nursing assistants, used QE to analyze the data, and compared their daily activities in real life with training experiences. A two-sample t-test with a large effect size (Cohen's d=1.63) indicated a significant difference between real-life and training activities. The findings suggested incorporating more empathetic training scenarios into the future design of our geriatric care simulation. The results have implications for human-computer interaction and human factors. This is illustrated by presenting an example of using QE to analyze expert interviews with nursing assistants as caregivers to inform subsequent design processes.

cs.HC↗

arXiv2024

Functional Near-Infrared Spectroscopy (fNIRS) Analysis of Interaction Techniques in Touchscreen-Based Educational Gaming

Educational games enhance learning experiences by integrating touchscreens, making interactions more engaging and intuitive for learners. However, the cognitive impacts of educational gameplay input modalities, such as the hand and stylus technique, are unclear. We compared the experience of using hands vs. a stylus for touchscreens while playing an educational game by analyzing oxygenated hemoglobin collected by functional Near-Infrared Spectroscopy and self-reported measures. In addition, we measured the hand vs. the stylus modalities of the task and calculated the relative neural efficiency and relative neural involvement using the mental demand and the quiz score. Our findings show that the hand condition had a significantly lower neural involvement, yet higher neural efficiency than the stylus condition. This result suggests the requirement of less cognitive effort while using the hand. Additionally, the self-reported measures show significant differences, and the results suggest that hand-based input is more intuitive, less cognitively demanding, and less frustrating. Conversely, the use of a stylus required higher cognitive effort due to the cognitive balance of controlling the pen and answering questions. These findings highlight the importance of designing educational games that allow learners to engage with the system while minimizing cognitive effort.

cs.HC↗

arXiv2024

Towards Anatomy Education with Generative AI-based Virtual Assistants in Immersive Virtual Reality Environments

Virtual reality (VR) and interactive 3D visualization systems have enhanced educational experiences and environments, particularly in complicated subjects such as anatomy education. VR-based systems surpass the potential limitations of traditional training approaches in facilitating interactive engagement among students. However, research on embodied virtual assistants that leverage generative artificial intelligence (AI) and verbal communication in the anatomy education context is underrepresented. In this work, we introduce a VR environment with a generative AI-embodied virtual assistant to support participants in responding to varying cognitive complexity anatomy questions and enable verbal communication. We assessed the technical efficacy and usability of the proposed environment in a pilot user study with 16 participants. We conducted a within-subject design for virtual assistant configuration (avatar- and screen-based), with two levels of cognitive complexity (knowledge- and analysis-based). The results reveal a significant difference in the scores obtained from knowledge- and analysis-based questions in relation to avatar configuration. Moreover, results provide insights into usability, cognitive task load, and the sense of presence in the proposed virtual assistant configurations. Our environment and results of the pilot study offer potential benefits and future research directions beyond medical education, using generative AI and embodied virtual agents as customized virtual conversational assistants.

cs.HC↗

arXiv2023

Cognitive Engagement for STEM+C Education: Investigating Serious Game Impact on Graph Structure Learning with fNIRS

For serious games on education, understanding the effectiveness of different learning methods in influencing cognitive processes remains a significant challenge. This study investigates the impact of serious games on graph structure learning. For this, we compared our in-house game-based learning (GBL) and video-based learning (VBL) methodologies by evaluating their effectiveness on cognitive processes by oxygenated hemoglobin levels using functional near-infrared spectroscopy (fNIRS). We conducted a 2 x 1 between subjects preliminary study with twelve participants, involving two conditions: game and video. Both groups received equivalent content related to the basic structure of a graph, with comparable session lengths. The game group interacted with a quiz-based game, while the video group watched a pre-recorded video. The fNIRS was employed to capture cerebral signals from the prefrontal cortex, and participants completed pre- and post- questionnaires capturing user experience and knowledge gain. In our study, we noted that the mean levels of oxygenated hemoglobin were higher in the GBL group, suggesting the potential enhanced cognitive involvement. Our results show that the lateral prefrontal cortex (LPFC) has greater hemodynamic activity during the learning period. Moreover, knowledge gain analysis showed an increase in mean score in the GBL group compared to the VBL group. Although we did not observe statistically significant changes due to participant variability and sample size, this preliminary work contributes to understanding how GBL and VBL impact cognitive processes, providing insights for enhanced instructional design and educational game development. Additionally, it emphasizes the necessity for further investigation into the impact of GBL on cognitive engagement and learning outcomes.

cs.HC↗