Engineering and Information Technology Collected Works - Research Publications
Permanent URI for this collection
Search Results
1 - 10 of 200
-
ItemCapturing prediction uncertainty in upstream cell culture models using conformal prediction and Gaussian processes(ML Research Press, 2023)This extended abstract compares the efficacy of Gaussian process and conformal XGBoost regressions in capturing prediction uncertainty in simulated and industrial cell culture data.
-
ItemAn Uncertainty-Accuracy-Based Score Function for Wrapper Methods in Feature Selection(IEEE, 2023-08-13)Feature Selection (FS) is an effective preprocessing method to deal with the curse of dimensionality in machine learning. Redundant features in datasets decrease the classification performance and increase the computational complexity. Wrapper methods are an important category of FS methods that evaluate various feature subsets and select the best one using performance measures related to a classifier. In these methods, the accuracy of classifiers is the most common performance measure for FS. Although the performance of classifiers depends on their uncertainty, this important criterion is neglected in these methods. In this paper, we present a new performance measure called Uncertainty-Accuracy-based Performance Measure for Feature Selection (UAPMFS) that uses an ensemble approach to measure both the accuracy and uncertainty of classifiers. UAPMFS uses bagging and uncertainty confusion matrix. This performance measure can be used in all wrapper methods to improve FS performance. We design two experiments to evaluate the performance of UAPMFS in wrapper methods. In experiments, we use the leave-one-variable-out strategy as the common strategy in wrapper methods to evaluate features. We also define a feature score function based on UAPMFS to rank and select features. In the first experiment, we investigate the importance of considering uncertainty in the FS process and show how neglecting uncertainty affects FS performance. In the second experiment, we compare the performance of the UAPMFS-based feature score function with the most common feature score functions for FS. Experimental results show the effectiveness of the proposed performance measure on different datasets.
-
ItemWork-In-Progress: Global Competency in Engineering Education - Are We Doing Well Enough?(Aalborg University Press, 2023)Global competence represents one’s ability to work effectively across cultures. It is a crucial employability skill for graduate mobility in the international engineering market and working in a diverse workforce. This study investigated the presence of global competence in an electronic engineering program’s curriculum at Universiti Tun Hussein Onn Malaysia (UTHM). Curriculum mapping was undertaken on the publicly accessible program handbooks and courses syllabus using a developed phrase bank of keywords indicative of global competence. Each course was then classified by the about matching to the keywords (strong, moderate, weak, no). A visual heat map was produced from the results to visually demonstrate the inclusion of global competence in courses. The findings demonstrated that 36%of the total courses in the electronic engineering program had strong or moderate evidence of global competency. This heatmap snapshot is a valuable reference for policymakers and curriculum developers to enable continual curriculum development works.
-
ItemNo Preview AvailableInvestigating Online Browsing Practices(ACM, 2023-03-19)
-
ItemProbing Power by Prompting: Harnessing Pre-trained Language Models for Power Connotation Framing(Association for Computational Linguistics, 2023)When describing actions, subtle changes in word choice can evoke very different associations with the involved entities. For instance, a company ‘employing workers’ evokes a more positive connotation than the one ‘exploiting’ them. This concept is called connotation. This paper investigates whether pre-trained language models (PLMs) encode such subtle connotative information about power differentials between involved entities. We design a probing framework for power connotation, building on (CITATION)’s operationalization of connotation frames. We show that zero-shot prompting of PLMs leads to above chance prediction of power connotation, however fine-tuning PLMs using our framework drastically improves their accuracy. Using our fine-tuned models, we present a case study of power dynamics in US news reporting on immigration, showing the potential of our framework as a tool for understanding subtle bias in the media.
-
ItemMATLAB Grader for Flexible Automated Assessment and Feedback in Large-Scale Engineering Subjects(AAEE, 2022)CONTEXT: One of the challenges in teaching large classes is the implementation of effective assessment strategies that deliver high quality, timely, and consistent feedback. This is particularly difficult in engineering subjects where assessment tasks involve programming concepts. It has previously been reported that automated assessment tools can be beneficial when there are large numbers of students. The introduction of automated assessment tools can significantly reduce the effort associated with manual marking, eliminate marking inconsistencies arising from having multiple graders, and most importantly, improve student self-learning by providing near immediate feedback. PURPOSE: The objective of this project was to evaluate the implementation of MATLAB Grader as an automated assessment and feedback tool in three engineering subjects across different year levels. Various assessment strategies in the deployment of MATLAB Grader to maximise students learning and engagement will be explored in this paper, including the integration of MATLAB Grader into the University's current Learning Management System (LMS), Canvas. APPROACH: Various deployment and grading strategies for MATLAB Grader were applied across three subjects: a first-year introductory engineering subject, a second-year biomedical engineering subject, and a Masters-level mechanical engineering subject. In the first-year subject, MATLAB programming is introduced as a basic problem-solving tool in the application of engineering concepts, while both the second-year and Masters-level subjects build on this foundation to explore more complex programming concepts and engineering applications. Student outcomes and responses were evaluated for MATLAB Grader assessments set up with unlimited attempts in the first and secondyear subjects, versus those set up with limited attempts with pre-tests in the Masters-level subject. OUTCOMES: Cohort-wide surveys on student experience with MATLAB Grader have yielded positive results in both the first and second-year subjects. Students generally appreciated the ease of access to assessment problems via the LMS and having unlimited attempts for their assessments. However, the quality of feedback provided has been identified as an area for improvement for both subjects. For the Masters-level subject, informal surveys and discussions with students also indicated feedback as a major area for improvement. The limited attempts with pre-tests strategy combined with minor technical issues have resulted in anxiety in a subset of the cohort when making submissions. CONCLUSION: MATLAB Grader offers flexibility in automated assessment and feedback across different disciplinespecific engineering subjects and various year levels. The deployment and effectiveness of the platform is highly dependent on the specific learning objectives and focus of a particular subject.
-
ItemImplementation of a Virtual Mechanics Laboratory for a first year undergraduate engineering subject using MATLAB App Designer(Australasian Association for Engineering Education, 2021)CONTEXT: In wake of the COVID-19 situation in 2020, when universities were faced with the transition from face-to-face learning to online delivery, many educators found themselves tasked with having to convert previously classroom-based teaching material to an equivalent online adaption. The transition was particularly challenging in a first-year engineering subject where hands-on laboratory experiments play an important part in the learning of basic mechanics principles as a foundation in engineering. Adaptation of physical hands-on experiments into the form of interactive virtual simulations was necessary to ensure students had an equally comprehensive laboratory experience in the online delivery mode. PURPOSE: This paper describes the development and implementation of a virtual laboratory for a set of mechanics experiments as an alternative to the physical hands-on laboratory. The interactive simulation application replicates the procedures of a physical mass-spring system investigation that applies two fundamental mechanics concepts, resultant forces and principle of moments. APPROACH: The virtual laboratory application is a user-friendly graphical user interface (GUI) integrated with a program code that models a physical spring system, developed in MATLAB App Designer. Key features of the application include animated outputs and virtual measurement tools that emulates the procedures of the actual experiment and MATLAB modelling that takes into account inconsistencies that may arise in real measurements. For deployment purposes, the simulation program in App Designer was compiled into a standalone executable and run using the MATLAB runtime environment. OUTCOMES: The virtual laboratory activity was successfully conducted during the online workshop classes in the first-year engineering subject at the University of Melbourne across a cohort of over 600 students. The simulation application in the virtual setting achieved similar learning outcomes as the experiments in the physical setting, but the activity was completed in significantly shorter times as compared to the expected physical hands-on. CONCLUSIONS AND RECOMMENDATIONS: The virtual experiments offered efficiencies over physical experiments in terms of minimising experimental procedure delays and allowing more focus on concepts and theories but unavoidably compromising other hands-on experience such as equipment set-up, calibration, real-world experimental observation, and troubleshooting. For a more comprehensive virtual laboratory experience, future work to model the virtual environment more accurately to represent real world behaviour is recommended.
-
ItemLearning in the Panopticon: Examining the Potential Impacts of AI Monitoring on Students(ACM, 2022-11-29)In a panopticon, people are intrusively monitored across all areas of their lives. AI monitoring has been ever more widely adopted in ed- ucation, with increasingly intrusive monitoring of students. These changes potentially create ethical harms, but current ethical dis- cussions predominantly focus on legal and governance issues. The concerns of the majority of users—namely students—are neglected. Overlooking students’ concerns further increases their vulnerabil- ity. We use a student-centred and speculative approach through the Story Completion Method (SCM) to explore how students would po- tentially respond to intrusive AI monitoring in a higher education setting. Our study included 71 participants who elaborated on the story stems we provided to them. Through a blending of thematic analysis coding and the techniques of developing grounded the- ory, we reveal that the common responses of students to extensive AI monitoring included impacts on personal psychology, changed behaviour, and cognition. There are likely major disruptions to personal autonomy, identity and educational relationships. If we are to avoid a future ‘big brother’ classroom, further investigations using HCI methods are critical to understanding how to protect students in AI-dominated learning.
-
ItemCompetencies of Cybersecurity Leaders: A Review and Research Agenda(Association for Information Systems, 2022)Increasingly, large organisations are turning to cybersecurity leaders such as chief information security officers (CISOs) to protect their information resources against attack. The role of the cybersecurity leader is distinct from other cybersecurity professionals in its need for strategy and collaboration, and distinct from other business leaders in its need to maintain situational awareness against active adversaries. Because the role is so new, however, organisations and educators continue to conceptualise it as a senior technological role rather than a strategic, business-oriented role. This representation leaves open a gap between what is viewed as ‘business’ and what is viewed as ‘IT’ – a gap that can leave organisations vulnerable to attack. In this systematic review, we examine the literature on cybersecurity leaders to develop a picture of the competencies required. Following analysis, we propose a preliminary matrix of competencies required for cybersecurity leaders. We conclude with an agenda for further research.
-
ItemInternational contextual reliability of the Empathy and Care Questionnaire: the Australian context(Australasian Association for Engineering Education, 2022)Context: Empathy has been identified as a key employability skill for professionals, underlying many skills and attributes anticipated as required by future engineers. The relevance and development of empathy and care as perceived by engineers has been quantitatively described in the USA and Germany using the Empathy and Care Questionnaire (ECQ). Cross—national variations in empathy have been reported outside of the engineering context. Purpose or Goal: Building on research from the US and Germany, the aim of this paper is to confirm the transferability of the factor structure of the ECQ (Hess, Strobel, Pan, et al., 2017) in the Australian context. Given the influence of national culture on the development and experience of empathy, establishing contextual reliability of the ECQ is important to ensure conceptual consistency before conducting data analysis. Approach or Methodology/Methods: This paper presents a confirmatory factor analysis (CFA) of the ECQ using data collected from engineers practicing in Australia. Engineers’ perceptions of empathy and care within Australian engineering practice were collected using an online version of the ECQ (N = 183). A CFA was conducted to establish transferability of the item structure to the Australian context. Outcomes: A marginal fit of the structure of Hess, Strobel, Pan, et al. (2017)‘s five-factor model was indicated. Modification through incorporation of covariance errors improved the fit. These errors were observed to be associated with the wording of the items. The ECQ structure comprising of factors: i) existence of empathy and care within engineering work and practice, ii) importance of empathy and care within engineering practice, iii) potential benefits of a greater inclusion of empathy and care into engineering, iv) the value of empathy and care in relational aspects of engineering work, and v) the extent to which empathy and care are considered learnable, was found to transfer to the Australian context. Additionally, the factor relating to learnability of empathy and care exhibited comparatively lower internal consistency than other factors. Conclusions/Recommendations/Summary: The ECQ can be used with confidence in the structure transferring from the US context, for describing perceptions of empathy and care of engineers practicing in Australia. There are opportunities for further research to explore national context, and to enhance the structure and internal reliability of the instrument.