Svetlana Yanushkevich
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Svetlana Yanushkevich has a total of 212 co-authors. Here are the top co-authors by number of shared publications:
Kenneth Lai
University of Calgary,Biometric Technologies Laboratory,Department of Electrical and Software Engineering,Canada
19 shared publications- Kenneth Lai
University of Calgary,Biometric Technologies Laboratory,Department of Electrical and Software Engineering,Canada
18 shared publications - Vlad P. Shmerko
University of Calgary
15 shared publications Vlad Shmerko
15 shared publicationsThomas Truong
Department of Electrical & Software Engineering, Biometric Technologies Laboratory, University of Calgary, Calgary, Canada
11 shared publicationsMarina Gavrilova
University of Calgary, Alberta, Canada
8 shared publications
Latest Publications
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (12) , 7979-7990
Augmenting Expert Elicitation With LLMs for Building Causal Models to Aid Decision Making
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (12) , 7979-7990
Augmenting Expert Elicitation With LLMs for Building Causal Models to Aid Decision Making
- Open Access
AI Tools, Especially Chatbots and Disabled People: Views of Critical Disability Studies Students
AI tools, including chatbots, have become increasingly mainstream and could pose many problems for disabled people and their allies. Such problems could include the lack of accessibility of AI tools for disabled people, including disabled activists and disabled people as digital citizens. Chatbot outputs also may be inaccurate or of low quality due to biased or insufficient training data and might reflect biased and low literacy about the lived reality of disabled people amongst the people involved in setting up the training. Therefore, problems could also include a potential decrease in the public’s literacy about the lived realities of disabled people when AI tools such as chatbots are used without critically analyzing their outputs or using inappropriate prompts for inquiry. Critical disability studies students see themselves as allies of disabled people and want to make a positive difference in their lives. AI tools, including chatbots, are an emerging area that allies of disabled people have to think about because of their increasing impact on the lived reality of disabled people. Using an online survey hosted on a Qualtrics platform, we examined the perspectives of 89 critical disability studies students on the usability and impact of AI tools, including chatbots, on disabled people in areas including the following: accessibility; what knowledge is available and what and whose knowledge is reflected in the output of AI tools and chatbots; the impact of AI tools and chatbots’ output on the literacy of the users about the lived reality of disabled people; and the impact on ability judgment problems disabled people experience, such as ableism and disablism. We performed a thematic analysis of two open-ended questions and frequency (%) analysis of the other questions. Our 89 participants outlined many problems. One set of problems covered accessibility barriers and design-process shortcomings of AI tools, including chatbots. Another set was linked to content-related issues, such as chatbots not reflecting the lived realities of disabled people, the use of biased language and portrayals of disabled people, and the limited usefulness of chatbot-provided information. These issues are not only problematic for disabled people but also for their allies. Participants gave many suggestions regarding the practical impacts of not having data on the lived reality of disabled people, as well as how chatbots could be used, or what has to happen, for them to challenge ableist power structures rather than reinforcing it. All our findings suggest the need for various actions whereby AI ethics, “responsible AI”, and “AI for good” principles could be used to decrease the problems noted, as could be the concepts of AI justice and of ability judgment justice and other ability judgment-linked concepts.
Open AccessAI Tools, Especially Chatbots and Disabled People: Views of Critical Disability Studies Students
AI tools, including chatbots, have become increasingly mainstream and could pose many problems for disabled people and their allies. Such problems could include the lack of accessibility of AI tools for disabled people, including disabled activists and disabled people as digital citizens. Chatbot outputs also may be inaccurate or of low quality due to biased or insufficient training data and might reflect biased and low literacy about the lived reality of disabled people amongst the people involved in setting up the training. Therefore, problems could also include a potential decrease in the public’s literacy about the lived realities of disabled people when AI tools such as chatbots are used without critically analyzing their outputs or using inappropriate prompts for inquiry. Critical disability studies students see themselves as allies of disabled people and want to make a positive difference in their lives. AI tools, including chatbots, are an emerging area that allies of disabled people have to think about because of their increasing impact on the lived reality of disabled people. Using an online survey hosted on a Qualtrics platform, we examined the perspectives of 89 critical disability studies students on the usability and impact of AI tools, including chatbots, on disabled people in areas including the following: accessibility; what knowledge is available and what and whose knowledge is reflected in the output of AI tools and chatbots; the impact of AI tools and chatbots’ output on the literacy of the users about the lived reality of disabled people; and the impact on ability judgment problems disabled people experience, such as ableism and disablism. We performed a thematic analysis of two open-ended questions and frequency (%) analysis of the other questions. Our 89 participants outlined many problems. One set of problems covered accessibility barriers and design-process shortcomings of AI tools, including chatbots. Another set was linked to content-related issues, such as chatbots not reflecting the lived realities of disabled people, the use of biased language and portrayals of disabled people, and the limited usefulness of chatbot-provided information. These issues are not only problematic for disabled people but also for their allies. Participants gave many suggestions regarding the practical impacts of not having data on the lived reality of disabled people, as well as how chatbots could be used, or what has to happen, for them to challenge ableist power structures rather than reinforcing it. All our findings suggest the need for various actions whereby AI ethics, “responsible AI”, and “AI for good” principles could be used to decrease the problems noted, as could be the concepts of AI justice and of ability judgment justice and other ability judgment-linked concepts.
- Open Access
- 17 September 2026
- Published by Elsevier in SSRN Electronic Journal
An Open-Source MAX86141-Based Platform for fNIRS and Optical Oxygenation Monitoring Research
Open-source functional near-infrared spectroscopy (fNIRS) platforms have improved access to wearable neuroimaging, but many existing designs rely on significant hand-assembly or include components that have reached end-of-life (EOL). This limits reproducibility and makes it difficult to adopt or extend prior designs. To address this, we developed a single-PCB fNIRS module built using modern components and designed for fabrication through standard PCBA services. The design costs approximately $42 per module to fabricate and assemble.To validate the design, breath-hold tests were performed to demonstrate the design produces physiologically plausible blood oxygenation measurements. Preliminary results indicate that the platform can support relative oxygenation monitoring during simple respiratory challenges, thus providing a practical foundation for future open-source fNIRS research.
- 17 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAn Open-Source MAX86141-Based Platform for fNIRS and Optical Oxygenation Monitoring Research
Open-source functional near-infrared spectroscopy (fNIRS) platforms have improved access to wearable neuroimaging, but many existing designs rely on significant hand-assembly or include components that have reached end-of-life (EOL). This limits reproducibility and makes it difficult to adopt or extend prior designs. To address this, we developed a single-PCB fNIRS module built using modern components and designed for fabrication through standard PCBA services. The design costs approximately $42 per module to fabricate and assemble.To validate the design, breath-hold tests were performed to demonstrate the design produces physiologically plausible blood oxygenation measurements. Preliminary results indicate that the platform can support relative oxygenation monitoring during simple respiratory challenges, thus providing a practical foundation for future open-source fNIRS research.
- Open Access
- 22 January 2026
- Published by Elsevier in Computer Vision and Image Understanding
- Vol. 265
Biometric technology roadmapping for personalized augmentative and alternative communication
- 22 January 2026
- Published by Elsevier in Computer Vision and Image Understanding
- Vol. 265
Open AccessBiometric technology roadmapping for personalized augmentative and alternative communication
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 147672-147684
Improving Stress Detection Using Doddington’s Menagerie
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 147672-147684
Open AccessImproving Stress Detection Using Doddington’s Menagerie
- 30 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Designing Stress Response Analyzer using Composite Cardiovascular Biomarkers
The number of processing channels is a critical requirement for wearable stress analyzers, especially for first responder needs (firefighters, police, rescuers, paramedics, and emergency medics). This study proposes reducing the number of processing channels in a human stress analyzer, which is a multi-channel wrist wearable wireless device, by introducing a composite approach that fuses cardiovascular biomarkers into a single processing stream. The proposed composite technique is justified experimentally for stress related tasks: accuracy rates ranging from 78% to 86.2% for detecting cognitive workload and 85.2% to 90.7% for detecting the psychophysical workload.
- 30 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Designing Stress Response Analyzer using Composite Cardiovascular Biomarkers
The number of processing channels is a critical requirement for wearable stress analyzers, especially for first responder needs (firefighters, police, rescuers, paramedics, and emergency medics). This study proposes reducing the number of processing channels in a human stress analyzer, which is a multi-channel wrist wearable wireless device, by introducing a composite approach that fuses cardiovascular biomarkers into a single processing stream. The proposed composite technique is justified experimentally for stress related tasks: accuracy rates ranging from 78% to 86.2% for detecting cognitive workload and 85.2% to 90.7% for detecting the psychophysical workload.
- 10 May 2025
- Published by Springer Nature
- p. 1283-1286
Inverse Biometrics: Privacy, Risks, and Trust
- 10 May 2025
- Published by Springer Nature
- p. 692-696
DNA
- Open Access
Probabilistic causal graphs as categorical data synthesizers: Do they do better than Gaussian Copulas and Conditional Tabular GANs?
This study investigates the generation of high-quality synthetic categorical data, such as survey data, using causal graph models. Generating synthetic data aims not only to create a variety of data for training the models but also to preserve privacy while capturing relationships between the data. The research employs Structural Equation Modeling (SEM) followed by Bayesian Networks (BN). We used the categorical data that are based on the survey of accessibility to services for people with disabilities. We created both SEM and BN models to represent causal relationships and to capture joint distributions between variables. In our case studies, such variables include, in particular, demographics, types of disability, types of accessibility barriers and frequencies of encountering those barriers. The study compared the SEM-based BN method with alternative approaches, including the probabilistic Gaussian copula technique and generative models like the Conditional Tabular Generative Adversarial Network (CTGAN). The proposed method outperformed others in statistical metrics, including the Chi-square test, Kullback-Leibler divergence, and Total Variation Distance (TVD). In particular, the BN model demonstrated superior performance, achieving the highest TVD, indicating alignment with the original data. The Gaussian Copula ranked second, while CTGAN exhibited moderate performance. These analyses confirmed the ability of the SEM-based BN to produce synthetic data that maintain statistical and relational validity while maintaining confidentiality. This approach is particularly beneficial for research on sensitive data, such as accessibility and disability studies.
Open AccessProbabilistic causal graphs as categorical data synthesizers: Do they do better than Gaussian Copulas and Conditional Tabular GANs?
This study investigates the generation of high-quality synthetic categorical data, such as survey data, using causal graph models. Generating synthetic data aims not only to create a variety of data for training the models but also to preserve privacy while capturing relationships between the data. The research employs Structural Equation Modeling (SEM) followed by Bayesian Networks (BN). We used the categorical data that are based on the survey of accessibility to services for people with disabilities. We created both SEM and BN models to represent causal relationships and to capture joint distributions between variables. In our case studies, such variables include, in particular, demographics, types of disability, types of accessibility barriers and frequencies of encountering those barriers. The study compared the SEM-based BN method with alternative approaches, including the probabilistic Gaussian copula technique and generative models like the Conditional Tabular Generative Adversarial Network (CTGAN). The proposed method outperformed others in statistical metrics, including the Chi-square test, Kullback-Leibler divergence, and Total Variation Distance (TVD). In particular, the BN model demonstrated superior performance, achieving the highest TVD, indicating alignment with the original data. The Gaussian Copula ranked second, while CTGAN exhibited moderate performance. These analyses confirmed the ability of the SEM-based BN to produce synthetic data that maintain statistical and relational validity while maintaining confidentiality. This approach is particularly beneficial for research on sensitive data, such as accessibility and disability studies.
- Open Access
Can LLMs Assist Expert Elicitation for Probabilistic Causal Modeling?
Objective: This study investigates the potential of Large Language Models (LLMs) as an alternative to human expert elicitation for extracting structured causal knowledge and facilitating causal modeling in biometric and healthcare applications. Material and Methods: LLM-generated causal structures, specifically Bayesian networks (BNs), were benchmarked against traditional statistical methods (e.g., Bayesian Information Criterion) using healthcare datasets. Validation techniques included structural equation modeling (SEM) to verifying relationships, and measures such as entropy, predictive accuracy, and robustness to compare network structures. Results and Discussion: LLM-generated BNs demonstrated lower entropy than expert-elicited and statistically generated BNs, suggesting higher confidence and precision in predictions. However, limitations such as contextual constraints, hallucinated dependencies, and potential biases inherited from training data require further investigation. Conclusion: LLMs represent a novel frontier in expert elicitation for probabilistic causal modeling, promising to improve transparency and reduce uncertainty in the decision-making using such models.
Open AccessCan LLMs Assist Expert Elicitation for Probabilistic Causal Modeling?
Objective: This study investigates the potential of Large Language Models (LLMs) as an alternative to human expert elicitation for extracting structured causal knowledge and facilitating causal modeling in biometric and healthcare applications. Material and Methods: LLM-generated causal structures, specifically Bayesian networks (BNs), were benchmarked against traditional statistical methods (e.g., Bayesian Information Criterion) using healthcare datasets. Validation techniques included structural equation modeling (SEM) to verifying relationships, and measures such as entropy, predictive accuracy, and robustness to compare network structures. Results and Discussion: LLM-generated BNs demonstrated lower entropy than expert-elicited and statistically generated BNs, suggesting higher confidence and precision in predictions. However, limitations such as contextual constraints, hallucinated dependencies, and potential biases inherited from training data require further investigation. Conclusion: LLMs represent a novel frontier in expert elicitation for probabilistic causal modeling, promising to improve transparency and reduce uncertainty in the decision-making using such models.