Umair Rehman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Umair Rehman has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
- Muhammad Umair Shah
Department of Management Sciences, Faculty of Engineering, University of Waterloo, Waterloo, Canada
10 shared publications - Amir Zaib Abbasi
King Fahd University of Petroleum and Minerals Center for Finance and Digital Economy Dhahran
8 shared publications - Muhammad Umair Danish
Department of Electrical and Computer Engineering, Western University
6 shared publications - Farkhund Iqbal
Zayed University
5 shared publications - Katarina Grolinger
Department of Electrical and Computer Engineering, Western University, London, Canada
5 shared publications - Alisa Grigorovich
Brock University
4 shared publications
Latest Publications
- Open Access
- 16 August 2026
- Published by Elsevier in SSRN Electronic Journal
Deep Mechanistic Explanations (DeepMex) to Unpack the Inner Operations of Memory Networks for Energy Forecasting
This paper introduces DeepMex (Deep Mechanistic Explanations), a multi-stage mechanistic interpretability approach designed to unpack the internal operations of memory network components in energy forecasting, enabling comprehensive explainability of model decisions. Existing interpretability methods for energy applications primarily rely on feature attribution techniques, such as SHAP or LIME, which merely quantify feature contributions to outputs but do not uncover internal decision-making pathways. To unveil inner operations of memory networks, we propose a DeepMex, which consist of four-stage mechanistic interpretability, where Stage 1 tracks feature influences and their interactions within model components; Stage 2 explains the roles of learned components and activations of each gate using activation patching; Stage 3 analyzes the effect of real-world input changes on forecasts, observes neuron bin shifts, and then extends to subpopulation analysis. Stage 4 goes further by examining neuron behavior at specific time steps to understand how the model behaves at a specific time of day. We conducted a theoretical analysis to ensure the correctness of the proposed method. Extensive experiments on ten diverse real-world energy datasets were conducted to explain the contributions of features, each gate, and the model's memory. The \deepmex provides actionable insights, fostering stakeholder trust in the network's forecasts.
- 16 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDeep Mechanistic Explanations (DeepMex) to Unpack the Inner Operations of Memory Networks for Energy Forecasting
This paper introduces DeepMex (Deep Mechanistic Explanations), a multi-stage mechanistic interpretability approach designed to unpack the internal operations of memory network components in energy forecasting, enabling comprehensive explainability of model decisions. Existing interpretability methods for energy applications primarily rely on feature attribution techniques, such as SHAP or LIME, which merely quantify feature contributions to outputs but do not uncover internal decision-making pathways. To unveil inner operations of memory networks, we propose a DeepMex, which consist of four-stage mechanistic interpretability, where Stage 1 tracks feature influences and their interactions within model components; Stage 2 explains the roles of learned components and activations of each gate using activation patching; Stage 3 analyzes the effect of real-world input changes on forecasts, observes neuron bin shifts, and then extends to subpopulation analysis. Stage 4 goes further by examining neuron behavior at specific time steps to understand how the model behaves at a specific time of day. We conducted a theoretical analysis to ensure the correctness of the proposed method. Extensive experiments on ten diverse real-world energy datasets were conducted to explain the contributions of features, each gate, and the model's memory. The \deepmex provides actionable insights, fostering stakeholder trust in the network's forecasts.
- 5 July 2026
- Published by Springer Nature
- p. 202-216
Can Multimodal LLMs Model Expert Ratings in Mobile UI Usability Evaluation?
- 5 July 2026
- Published by Springer Nature
- p. 202-216
Can Multimodal LLMs Model Expert Ratings in Mobile UI Usability Evaluation?
- Open Access
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Applied Perception
- Vol. 23 (1) , 1-32
Decoding the Connection: Viewer Experience and Video Quality through Human-Centered Constructs
This research explores the impact of video quality on viewer experience (VX) in the digital age. Videos are ubiquitous in our lives, yet our understanding of how quality variations affect satisfaction and engagement remains limited. By introducing a one-to-many relationship between Quality of Service (QoS) and Quality of Experience (QoE), the study aims to provide practical and deeper insights for content creators and streaming platforms that contemporary subjective metrics cannot provide. It introduces the concept of VX, a novel extension of the QoE, to better capture the complexities of human response to multimedia content. The research combines qualitative and quantitative methods, utilizing established quality assessment frameworks like SSIMplus. Through a combination of statistical and thematic explorations, we provide the basis of a novel framework that has real-world implications for enhancing user satisfaction and the overall quality of video-based content in an increasingly digital world.
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Applied Perception
- Vol. 23 (1) , 1-32
Open AccessDecoding the Connection: Viewer Experience and Video Quality through Human-Centered Constructs
This research explores the impact of video quality on viewer experience (VX) in the digital age. Videos are ubiquitous in our lives, yet our understanding of how quality variations affect satisfaction and engagement remains limited. By introducing a one-to-many relationship between Quality of Service (QoS) and Quality of Experience (QoE), the study aims to provide practical and deeper insights for content creators and streaming platforms that contemporary subjective metrics cannot provide. It introduces the concept of VX, a novel extension of the QoE, to better capture the complexities of human response to multimedia content. The research combines qualitative and quantitative methods, utilizing established quality assessment frameworks like SSIMplus. Through a combination of statistical and thematic explorations, we provide the basis of a novel framework that has real-world implications for enhancing user satisfaction and the overall quality of video-based content in an increasingly digital world.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems, Man, and Cybernetics Letters
- Vol. 1, 71-77
Augmented Reality as a Training-System Scaffold: Evidence From a Real-Aircraft Study With Certified Flight Instructors
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems, Man, and Cybernetics Letters
- Vol. 1, 71-77
Open AccessAugmented Reality as a Training-System Scaffold: Evidence From a Real-Aircraft Study With Certified Flight Instructors
- Open Access
Seeing Through Risk: A Symbolic Approximation of Prospect Theory
We propose a novel symbolic modeling framework for decision-making under risk that merges interpretability with the core insights of Prospect Theory. Our approach replaces opaque utility curves and probability weighting functions with transparent, effect-size-guided features. We mathematically formalize the method, demonstrate its ability to replicate well-known framing and loss-aversion phenomena, and provide an end-to-end empirical validation on synthetic datasets. The resulting model achieves competitive predictive performance while yielding clear coefficients mapped onto psychological constructs, making it suitable for applications ranging from AI safety to economic policy analysis.
Open AccessSeeing Through Risk: A Symbolic Approximation of Prospect Theory
We propose a novel symbolic modeling framework for decision-making under risk that merges interpretability with the core insights of Prospect Theory. Our approach replaces opaque utility curves and probability weighting functions with transparent, effect-size-guided features. We mathematically formalize the method, demonstrate its ability to replicate well-known framing and loss-aversion phenomena, and provide an end-to-end empirical validation on synthetic datasets. The resulting model achieves competitive predictive performance while yielding clear coefficients mapped onto psychological constructs, making it suitable for applications ranging from AI safety to economic policy analysis.
- 4 April 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (3) , 1-35
Towards a unified evaluation framework: integrating human perception and metrics for AI-generated images
- 4 April 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (3) , 1-35
Towards a unified evaluation framework: integrating human perception and metrics for AI-generated images
- Open Access
- 28 February 2025
- Published by Springer Nature in Research Square
AI-SSIM: Human-Centric Image Assessment through Pseudo-Reference Generation and Logical Consistency Analysis in AI-Generated Visuals
We present AI-SSIM, a computational image metric for assessing the quality and logical consistency of AI-generated and real-world images. Traditional metrics like structural similarity index measure (SSIM) and multi-scale structural similarity index measure (MS-SSIM) require a ground-truth image, which is often unavailable in AI-generated imagery, and overlook key factors such as logical coherence and content usability. AI-SSIM addresses these gaps by employing advanced pre-trained models to generate a pseudo-reference image, convolution and attention layers to evaluate image quality, and adaptive pooling to minimize distortion during resizing pseudo-reference images. We also designed and statistically validated a multi-item questionnaire for assessing image quality. AI-SSIM was benchmarked against human scales and compared to both full-reference and no-reference metrics, where it demonstrated superior accuracy. The proposed metric has broad applicability, as it can compute scores in both scenarios where ground-truth images are either available or absent.
- 28 February 2025
- Published by Springer Nature in Research Square
Open AccessAI-SSIM: Human-Centric Image Assessment through Pseudo-Reference Generation and Logical Consistency Analysis in AI-Generated Visuals
We present AI-SSIM, a computational image metric for assessing the quality and logical consistency of AI-generated and real-world images. Traditional metrics like structural similarity index measure (SSIM) and multi-scale structural similarity index measure (MS-SSIM) require a ground-truth image, which is often unavailable in AI-generated imagery, and overlook key factors such as logical coherence and content usability. AI-SSIM addresses these gaps by employing advanced pre-trained models to generate a pseudo-reference image, convolution and attention layers to evaluate image quality, and adaptive pooling to minimize distortion during resizing pseudo-reference images. We also designed and statistically validated a multi-item questionnaire for assessing image quality. AI-SSIM was benchmarked against human scales and compared to both full-reference and no-reference metrics, where it demonstrated superior accuracy. The proposed metric has broad applicability, as it can compute scores in both scenarios where ground-truth images are either available or absent.
- Open Access
Leveraging Order-Theoretic Tournament Graphs for Assessing Internal Consistency in Survey-Based Instruments Across Diverse Scenarios
This paper introduces Monotone Delta, an order-theoretic measure designed to enhance the reliability assessment of survey-based instruments in human-machine interactions. Traditional reliability measures, such as Cronbach's Alpha and McDonald's Omega, often yield misleading estimates due to their sensitivity to redundancy, multidimensional constructs, and assumptions of normality and uncorrelated errors. These limitations can compromise decision-making in human-centric evaluations, where survey instruments inform adaptive interfaces, cognitive workload assessments, and human-AI trust models. Monotone Delta addresses these issues by quantifying internal consistency through the minimization of ordinal contradictions and alignment with a unidimensional latent order using weighted tournaments. Unlike traditional approaches, it operates without parametric or model-based assumptions. We conducted theoretical analyses and experimental evaluations on four challenging scenarios: tau-equivalence, redundancy, multidimensionality, and non-normal distributions, and proved that Monotone Delta provides more stable reliability assessments compared to existing methods. The Monotone Delta is a valuable alternative for evaluating questionnaire-based assessments in psychology, human factors, healthcare, and interactive system design, enabling organizations to optimize survey instruments, reduce costly redundancies, and enhance confidence in human-system interactions.
Open AccessLeveraging Order-Theoretic Tournament Graphs for Assessing Internal Consistency in Survey-Based Instruments Across Diverse Scenarios
This paper introduces Monotone Delta, an order-theoretic measure designed to enhance the reliability assessment of survey-based instruments in human-machine interactions. Traditional reliability measures, such as Cronbach's Alpha and McDonald's Omega, often yield misleading estimates due to their sensitivity to redundancy, multidimensional constructs, and assumptions of normality and uncorrelated errors. These limitations can compromise decision-making in human-centric evaluations, where survey instruments inform adaptive interfaces, cognitive workload assessments, and human-AI trust models. Monotone Delta addresses these issues by quantifying internal consistency through the minimization of ordinal contradictions and alignment with a unidimensional latent order using weighted tournaments. Unlike traditional approaches, it operates without parametric or model-based assumptions. We conducted theoretical analyses and experimental evaluations on four challenging scenarios: tau-equivalence, redundancy, multidimensionality, and non-normal distributions, and proved that Monotone Delta provides more stable reliability assessments compared to existing methods. The Monotone Delta is a valuable alternative for evaluating questionnaire-based assessments in psychology, human factors, healthcare, and interactive system design, enabling organizations to optimize survey instruments, reduce costly redundancies, and enhance confidence in human-system interactions.
- 3 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 55 (2) , 223-233
Global-Local Image Perceptual Score (GLIPS): Evaluating Photorealistic Quality of AI-Generated Images
This article introduces the global-local image perceptual score (GLIPS), an image metric designed to assess the photorealistic image quality of AI-generated images with a high degree of alignment to human visual perception. Traditional metrics such as Fréchet inception distance (FID) and kernel inception distance scores do not align closely with human evaluations. The proposed metric incorporates advanced transformer-based attention mechanisms to assess local similarity and maximum mean discrepancy to evaluate global distributional similarity. To evaluate the performance of GLIPS, we conducted a human study on photorealistic image quality. Comprehensive tests across various generative models demonstrate that GLIPS consistently outperforms existing metrics like FID, structural similarity index measure, and multiscale structural similarity index measure in terms of correlation with human scores. In addition, we introduce the interpolative binning scale, a refined scaling method that enhances the interpretability of metric scores by aligning them more closely with human evaluative standards. The proposed metric and scaling approach not only provide more reliable assessments of AI-generated images but also suggest pathways for future enhancements in image generation technologies.
- 3 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 55 (2) , 223-233
Global-Local Image Perceptual Score (GLIPS): Evaluating Photorealistic Quality of AI-Generated Images
This article introduces the global-local image perceptual score (GLIPS), an image metric designed to assess the photorealistic image quality of AI-generated images with a high degree of alignment to human visual perception. Traditional metrics such as Fréchet inception distance (FID) and kernel inception distance scores do not align closely with human evaluations. The proposed metric incorporates advanced transformer-based attention mechanisms to assess local similarity and maximum mean discrepancy to evaluate global distributional similarity. To evaluate the performance of GLIPS, we conducted a human study on photorealistic image quality. Comprehensive tests across various generative models demonstrate that GLIPS consistently outperforms existing metrics like FID, structural similarity index measure, and multiscale structural similarity index measure in terms of correlation with human scores. In addition, we introduce the interpolative binning scale, a refined scaling method that enhances the interpretability of metric scores by aligning them more closely with human evaluative standards. The proposed metric and scaling approach not only provide more reliable assessments of AI-generated images but also suggest pathways for future enhancements in image generation technologies.
- 29 October 2024
- Published by Springer Nature in AI and Ethics
- Vol. 5 (3) , 2867-2880
Comparative analysis of moral decision-making and trust dynamics: human reasoning vs. ChatGPT-3 narratives
- 29 October 2024
- Published by Springer Nature in AI and Ethics
- Vol. 5 (3) , 2867-2880
Comparative analysis of moral decision-making and trust dynamics: human reasoning vs. ChatGPT-3 narratives