Research Direction
Human-centred AI for cognitive load, stress, and adaptive learning.
My research focuses on human-centred AI, physiological computing, affective computing, and adaptive learning systems. I investigate how physiological, behavioural, and learning-related signals can be used to understand cognitive load, stress, and individual differences, with the goal of developing intelligent systems that are interpretable, inclusive, practical, and ethically deployable.
Current Research Focus
My current research examines how physiological and behavioural signals can be used to understand stress and cognitive demands during learning. In recent work published at ACM ITiCSE 2026, my co-authors and I investigated physiological responses to acute socially evaluative stress during an introductory HTML learning activity using electrocardiogram and galvanic skin response signals.
I am particularly interested in translating these insights into adaptive learning systems that can adjust task difficulty, pacing, feedback, scaffolding, or recovery intervals while accounting for individual differences and avoiding additional cognitive burden.
🧠 + ⌚ + 🎓
Research Question
How can adaptive AI systems support learning while respecting human cognitive effort, stress, privacy, and individual differences?
Research Areas at a Glance
🧠
Cognitive Load and Stress
Understanding how mental effort and stress change across tasks, difficulty, confidence, and learning performance.
⌚
Physiological Computing
Analysing signals such as ECG, heart rate variability, and galvanic skin response to study stress, arousal, and recovery.
🎓
Adaptive Learning
Designing learning systems that adapt pacing, feedback, review strategy, scaffolding, and difficulty based on learner state.
Research Themes
Cognitive Load and Stress Modelling
Understanding how physiological, behavioural, and self-reported signals can be used to estimate cognitive load, stress, arousal, and mental effort during learning or work tasks.
Personalisation and Individual Differences
Studying how AI systems can adapt to individual baselines, learning patterns, perceived difficulty, confidence, and stress responses without requiring excessive amounts of personal data.
Lightweight Adaptive Learning
Designing adaptive learning tools that can recommend pacing, review strategy, task difficulty, instructional support, and recovery intervals using transparent and practical approaches.
Affective Computing for Education
Investigating how computational systems can recognise and respond to stress, arousal, confidence, and other learner states while preserving human agency and avoiding intrusive intervention.
Human-Centred AI Evaluation
Evaluating AI systems using measures beyond predictive accuracy, including cognitive load, trust, usability, calibration, fairness, practical usefulness, and learner wellbeing.
Responsible Physiological AI
Exploring privacy-aware, interpretable, and ethically deployable methods for working with sensitive physiological and behavioural data in educational and user-facing systems.
Research Impact
My research aims to connect artificial intelligence, human-computer interaction, physiological computing, and education. Rather than evaluating learning systems only through performance or accuracy, I am interested in how technology affects learners’ cognitive effort, stress, confidence, engagement, and capacity to recover from demanding tasks.
Understand Learner State
Develop evidence-based methods for understanding cognitive load, stress, arousal, and recovery during learning.
Support Adaptive Education
Inform systems that adapt pacing, task difficulty, feedback, hints, worked examples, or structured breaks when learners need support.
Promote Inclusive Design
Design learning technologies that recognise individual differences without treating one physiological or behavioural pattern as universal.
Publications and Manuscripts
Peer-Reviewed ACM Conference Publication
Impacts of Stress During CS Learning: Physiological Insights for Adaptive Education
Authors: Maliha Mian, Muhammad Arkaan Izhraqi, Nadine Marcus, and Gelareh Mohammadi
Venue: Proceedings of the 31st ACM Conference on Innovation and Technology in Computer Science Education, Volume 1 (ITiCSE 2026), Madrid, Spain
Publication: July 2026 · 7 pages
Research areas: Computer science education, affective computing, physiological computing, stress induction, cognitive load, gender differences, and adaptive learning.
Summary: This study investigated how acute socially evaluative stress manifests physiologically during an introductory HTML learning activity. Electrocardiogram and galvanic skin response signals were recorded to examine changes in heart activity, heart rate variability, skin conductance, and physiological arousal during and after a controlled stress task.
Key finding: The stress induction produced significant physiological responses that continued into subsequent learning activities. The study also identified different physiological response and recovery patterns across the participant gender groups, highlighting the importance of individual differences when designing inclusive and responsive learning systems.
Implications: The findings support future research on stress-aware educational technologies that could adapt task difficulty, pacing, feedback, scaffolding, or recovery support using physiological and behavioural indicators.
Manuscript / Preprint
Exploring Gender Differences in Cognitive Load and Stress Responses: A Machine Learning Approach Using Physiological Data in Learning Environments
Status: Manuscript in preparation / under review
Research areas: Machine learning, cognitive load, stress modelling, physiological signal analysis, individual differences, and educational technology.
Role: Methodology, implementation, experiments, physiological data analysis, machine-learning analysis, and initial manuscript drafting.
Peer-Reviewed Conference Publication
Improvement on Finger Region Extraction for Hand-Waving Finger Vein Authentication
Authors: Hiroyuki Suzuki, Muhammad Arkaan Izhraqi, Jumpei Nagata, Takashi Obi, and Takashi Komuro
Venue: JSAP–OSA Joint Symposia 2019, Sapporo, Japan
Publication: September 2019 · Conference paper 18p-E215-9
Research areas: Biometrics, computer vision, image processing, machine learning, and contactless finger-vein authentication.
Summary: This research investigated improved extraction of finger regions from moving hand images for a hand-waving finger-vein authentication system. Reliable finger-region extraction is an important preprocessing step for detecting vein patterns and supporting contactless biometric authentication while the user moves their hand in front of the imaging system.
Research contribution: The work contributes to the development of more practical contactless biometric systems by improving the computer-vision pipeline used to locate and isolate finger regions before finger-vein pattern analysis and identity verification.
Research Roadmap
A long-term pathway from research development and reproducible experimentation toward doctoral study in human-centred AI.
2026
ACM publication, research foundation, public notes, and lightweight prototypes.
2027
Prototype, benchmark, and reproducible experiments.
2028
PhD proposal, supervisor mapping, and research statement.
2029
Applications, outreach, writing sample, and final portfolio.
2030
Begin doctoral research in human-centred AI.
Open Collaboration
I am open to collaboration with researchers and practitioners working on human-centred AI, physiological computing, learning analytics, adaptive learning, affective computing, computer science education, and responsible AI systems.
I am particularly interested in interdisciplinary collaborations that connect technical development with meaningful human outcomes.
Arkaan Izhraqi 