Vivekanandan Suresh Kumar
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
Vivekanandan Suresh Kumar has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
Kinshuk
University of North Texas
20 shared publicationsDavid Boulanger
Athabasca University
13 shared publicationsKannan Govindarajan
School of Computing and Information Systems, Athabasca University, Edmonton, AB, Canada
8 shared publicationsColin Pinnell
Athabasca University
6 shared publicationsThamarai Selvi Somasundaram
Anna University, Chennai, India
6 shared publicationsJeremie Seanosky
Athabasca University
5 shared publications
Latest Publications
- 1 May 2024
- Published by Taylor & Francis
- p. 445-468
Explainable AI and AWE
Automated writing evaluation (AWE) is a complex research topic having a long way to go before becoming a mature and trustworthy technology. Research focuses on reporting the performance of AWE systems by measuring their predictive accuracy without analyzing what the algorithm actually learns. This lack of transparency is partly due to the high complexity and dimensionality of AWE, and the usage of deep learning to cope with them just exacerbated the problem. Interestingly, the ascent of explainable artificial intelligence (XAI) allows to look retrospectively at the way AWE is developed. However, XAI is still in its infancy facing many limitations. Consequently, it is crucial to understand these limitations and how to correctly interpret black-box scoring models. This chapter investigates the extent to which XAI can help determine the most generalizable model and its explanations help fix learned fallacies. The chapter presents several applications and cautions about XAI in the AWE field, drawing conclusions that (1) the explanation models of the random forests had much higher descriptive accuracy than those of the gradient-boosted trees, two inherently opaque architectures, and (2) the simplest AWE models produced the most stable, consistent, and interpretable explanations.
- 1 May 2024
- Published by Taylor & Francis
- p. 445-468
Explainable AI and AWE
Automated writing evaluation (AWE) is a complex research topic having a long way to go before becoming a mature and trustworthy technology. Research focuses on reporting the performance of AWE systems by measuring their predictive accuracy without analyzing what the algorithm actually learns. This lack of transparency is partly due to the high complexity and dimensionality of AWE, and the usage of deep learning to cope with them just exacerbated the problem. Interestingly, the ascent of explainable artificial intelligence (XAI) allows to look retrospectively at the way AWE is developed. However, XAI is still in its infancy facing many limitations. Consequently, it is crucial to understand these limitations and how to correctly interpret black-box scoring models. This chapter investigates the extent to which XAI can help determine the most generalizable model and its explanations help fix learned fallacies. The chapter presents several applications and cautions about XAI in the AWE field, drawing conclusions that (1) the explanation models of the random forests had much higher descriptive accuracy than those of the gradient-boosted trees, two inherently opaque architectures, and (2) the simplest AWE models produced the most stable, consistent, and interpretable explanations.
- Open Access
- 24 October 2023
- Published by IGI Global Scientific Publishing
- p. 1-20
Using Causal Inference in Learning Processes to Predict Student Proficiency
The search for causal relations from observational data is an open problem that spans many fields. In the area of learning, this is especially important. The ability to determine the effect of a new teaching strategy or the cause of an upswing in student performance is always desirable. In computer science, integrated development environments (IDE) offer students many features promising to instill the necessary competency skills for migration to industry. In this chapter, current causal discovery methods are applied to investigate a causal link between IDE consultations and student competency which is measured by the number of issues at the end of the coding timeline. The coding activities of students were timestamped over the coding lifetime. Due to the nature of the data, the authors were able to test for causality using methods for static and methods for dynamic data. The authors show the presence of a causal link between IDE consults and student improvement. In addition, they show the time it takes to see the effect of a system consult.
- 24 October 2023
- Published by IGI Global Scientific Publishing
- p. 1-20
Open AccessUsing Causal Inference in Learning Processes to Predict Student Proficiency
The search for causal relations from observational data is an open problem that spans many fields. In the area of learning, this is especially important. The ability to determine the effect of a new teaching strategy or the cause of an upswing in student performance is always desirable. In computer science, integrated development environments (IDE) offer students many features promising to instill the necessary competency skills for migration to industry. In this chapter, current causal discovery methods are applied to investigate a causal link between IDE consultations and student competency which is measured by the number of issues at the end of the coding timeline. The coding activities of students were timestamped over the coding lifetime. Due to the nature of the data, the authors were able to test for causality using methods for static and methods for dynamic data. The authors show the presence of a causal link between IDE consults and student improvement. In addition, they show the time it takes to see the effect of a system consult.
- Open Access
- 29 November 2022
- Published by University of Alberta Libraries in Canadian Journal of Learning and Technology
- Vol. 48 (4)
Cognification in Learning, Teaching, and Training
Over the past decade, opportunities for online learning have dramatically increased. Learners around the world now have digital access to a wide array of corporate trainings, certifications, comprehensive academic degree programs, and other educational and training options. Some organizations are blending traditional instruction methods with online technologies. Blended learning generates large volumes of data about both the content (quality and usage) and the learners (study habits and learning outcomes). Correspondingly, the need to properly process voluminous, continuous, and often disparate data has prompted the advent of cognification. Cognification techniques design complex data analytic models that allow natural intelligence to engage artificial smartness in ways that can enhance the learning experience. Cognification is the approach to make something increasingly, ethically, and regulatably smarter. This article highlights how emerging trends in cognification could disrupt online education.
- 29 November 2022
- Published by University of Alberta Libraries in Canadian Journal of Learning and Technology
- Vol. 48 (4)
Open AccessCognification in Learning, Teaching, and Training
Over the past decade, opportunities for online learning have dramatically increased. Learners around the world now have digital access to a wide array of corporate trainings, certifications, comprehensive academic degree programs, and other educational and training options. Some organizations are blending traditional instruction methods with online technologies. Blended learning generates large volumes of data about both the content (quality and usage) and the learners (study habits and learning outcomes). Correspondingly, the need to properly process voluminous, continuous, and often disparate data has prompted the advent of cognification. Cognification techniques design complex data analytic models that allow natural intelligence to engage artificial smartness in ways that can enhance the learning experience. Cognification is the approach to make something increasingly, ethically, and regulatably smarter. This article highlights how emerging trends in cognification could disrupt online education.
- Top Cited PapersOpen Access
- 15 September 2020
- Published by Elsevier in International Journal of Artificial Intelligence in Education
- Vol. 31 (3) , 538-584
Automated Essay Scoring and the Deep Learning Black Box: How Are Rubric Scores Determined?
- 15 September 2020
- Published by Elsevier in International Journal of Artificial Intelligence in Education
- Vol. 31 (3) , 538-584
Top Cited PapersOpen AccessAutomated Essay Scoring and the Deep Learning Black Box: How Are Rubric Scores Determined?
- Open Access
- 10 August 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 148052-148065
Visualizing Benefits: Evaluating Healthcare Information System Using IS-Impact Model
Reducing costs and optimizing operations are major challenges in many large-scale organizations including healthcare authorities. Research shows that despite ongoing investments in healthcare information systems (IS), the promised benefits often are partially realized or not at all. Part of the solution is to ensure that evaluation methodologies are available to clearly identify the success of these initiatives and from here articulate and mitigate the deficiencies or move to an alternative technology. The literature asserts that few practitioners have implemented a standardized evaluation approach. Using an established model, namely Information System Impact (IS-impact) model, we proposed a modified evaluation model to assess and a visualization tool to visualize the success of an information systems from a healthcare perspective. The modified IS-Impact model includes six constructs – individual impact, organization impact, provincial alignment impact, system quality, information quality, and service quality. We applied the modified IS-impact model and the proposed visualization tool against an existing healthcare software solution. An empirical study was conducted at a healthcare authority, with responses from 150 participants who use the healthcare IS, which confirmed that the proposed model and the visualization tool are valid and reliable to measure healthcare systems success. The evaluation model and the visualization tool are found to be efficient to narrow down the scope of inquiry from the general to the specific and quickly identifying the gaps and successes within the established software solution for the healthcare authority. Healthcare or clinical informatics researchers will be benefited from this research in evaluating the ongoing or nearly established healthcare information systems.
- 10 August 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 148052-148065
Open AccessVisualizing Benefits: Evaluating Healthcare Information System Using IS-Impact Model
Reducing costs and optimizing operations are major challenges in many large-scale organizations including healthcare authorities. Research shows that despite ongoing investments in healthcare information systems (IS), the promised benefits often are partially realized or not at all. Part of the solution is to ensure that evaluation methodologies are available to clearly identify the success of these initiatives and from here articulate and mitigate the deficiencies or move to an alternative technology. The literature asserts that few practitioners have implemented a standardized evaluation approach. Using an established model, namely Information System Impact (IS-impact) model, we proposed a modified evaluation model to assess and a visualization tool to visualize the success of an information systems from a healthcare perspective. The modified IS-Impact model includes six constructs – individual impact, organization impact, provincial alignment impact, system quality, information quality, and service quality. We applied the modified IS-impact model and the proposed visualization tool against an existing healthcare software solution. An empirical study was conducted at a healthcare authority, with responses from 150 participants who use the healthcare IS, which confirmed that the proposed model and the visualization tool are valid and reliable to measure healthcare systems success. The evaluation model and the visualization tool are found to be efficient to narrow down the scope of inquiry from the general to the specific and quickly identifying the gaps and successes within the established software solution for the healthcare authority. Healthcare or clinical informatics researchers will be benefited from this research in evaluating the ongoing or nearly established healthcare information systems.
- 25 May 2019
- Published by Springer Nature
- p. 123-151
Performance Analysis of a Serial Natural Language Processing Pipeline for Scaling Analytics of Academic Writing Process
- 25 May 2019
- Published by Springer Nature
- p. 123-151
Performance Analysis of a Serial Natural Language Processing Pipeline for Scaling Analytics of Academic Writing Process
- 4 November 2018
- Published by Springer Nature
- p. 49-67
Learning Traces, Competence Assessment, and Causal Inference for English Composition
- 4 November 2018
- Published by Springer Nature
- p. 49-67
Learning Traces, Competence Assessment, and Causal Inference for English Composition
- Open Access
- 25 August 2018
- Published by Springer Nature in Smart Learning Environments
- Vol. 5 (1) , 13
The Objective Ear: assessing the progress of a music task
- 25 August 2018
- Published by Springer Nature in Smart Learning Environments
- Vol. 5 (1) , 13
Open AccessThe Objective Ear: assessing the progress of a music task
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 188-190
Assessing Learning Analytics Systems Impact by Summative Measures
This paper introduces a randomized study conducted among a group of 48 student Java programmers to assess the impact of learning analytics (LA) on their academic performance. The LA system design incorporated both cognitive and metacognitive tools to help learners take possession of their learning processes. Participation was voluntary and data about potential confounding factors were also collected to minimize bias by blocking on two or more factors (future work). This paper summarily explores the relationships between students' programming expertise, coding assignments, user experience and satisfaction, and academic performance. The results of this preliminary exploration are inconclusive as to whether the LA system made a difference in academic performance. Nevertheless, they seem to indicate that LA was beneficial to student programmers and that summative measures such as grades are not a proper metric to measure the usefulness of LA systems.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 188-190
Assessing Learning Analytics Systems Impact by Summative Measures
This paper introduces a randomized study conducted among a group of 48 student Java programmers to assess the impact of learning analytics (LA) on their academic performance. The LA system design incorporated both cognitive and metacognitive tools to help learners take possession of their learning processes. Participation was voluntary and data about potential confounding factors were also collected to minimize bias by blocking on two or more factors (future work). This paper summarily explores the relationships between students' programming expertise, coding assignments, user experience and satisfaction, and academic performance. The results of this preliminary exploration are inconclusive as to whether the LA system made a difference in academic performance. Nevertheless, they seem to indicate that LA was beneficial to student programmers and that summative measures such as grades are not a proper metric to measure the usefulness of LA systems.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 199-201
Identifying Student Difficulty in a Digital Learning Environment
This paper discusses the development of TutorAlert, a natural language processing system similar to those used in sentiment analysis, but applied to the data generated by students in a digital online learning environment in order to detect confused or frustrated students. A number of machine learning algorithms were tested in the development process, including Support Vector Machines (SVM), Naive Bayes, and Random Forest classifiers. As well, an array of natural language preparation techniques were employed to determine the optimum preprocessing configuration to produce relevant results. We found that detecting potential student frustration or confusion was most successful using a Sequential Minimal Optimization algorithm (SMO), along with the Stanford Part-Of-Speech Tagger (POS Tagger), the iterated version of the Lovins stemmer, and a custom dictionary to help determine relevance probability. This model produced a promising initial F1 score of 0.79 and an accuracy of 0.83. Further, agreement values of 88% were achieved during inter-rater reliability testing between the classifier and human judges.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 199-201
Identifying Student Difficulty in a Digital Learning Environment
This paper discusses the development of TutorAlert, a natural language processing system similar to those used in sentiment analysis, but applied to the data generated by students in a digital online learning environment in order to detect confused or frustrated students. A number of machine learning algorithms were tested in the development process, including Support Vector Machines (SVM), Naive Bayes, and Random Forest classifiers. As well, an array of natural language preparation techniques were employed to determine the optimum preprocessing configuration to produce relevant results. We found that detecting potential student frustration or confusion was most successful using a Sequential Minimal Optimization algorithm (SMO), along with the Stanford Part-Of-Speech Tagger (POS Tagger), the iterated version of the Lovins stemmer, and a custom dictionary to help determine relevance probability. This model produced a promising initial F1 score of 0.79 and an accuracy of 0.83. Further, agreement values of 88% were achieved during inter-rater reliability testing between the classifier and human judges.