We are responding to calls for students to be trained in using generative AI (GAI). But training must take account of what students know; their prevailing attitudes; and their ethics with respect to using GAI. And knowing what our students know about GAI gives instructors an opportunity to co-construct meaningful teaching and learning moments. A class of 103 students, enrolled in the course Communication in the Engineering Sciences, in our BSc program in Electrical and Computer Engineering, was probed on their use of generative AI. The study draws on two datasets—student responses to a case study on unethical use of GAI and a questionnaire gathering qualitative data. Questionnaire responses reveal that generally students understand the basics of how GAI works. But this understanding is flawed, for example some students erroneously believe that GAI draws in real time from the Internet. Most respondents have used GAI in their studies, particularly to clarify concepts and summarize. We did not find widespread use of GAI for higher cognitive tasks. But this we suspect is linked to program sequencing: as students advance in their program more sophisticated uses of GAI are likely. This supports the need for longitudinal studies which track use in relation to program advancement. Like peers elsewhere, this class expressed concern about GAI's ability to propagate misinformation and bias; GAI-facilitated plagiarism; data privacy; and overreliance leading to impairment of learning. As it relates to whether GAI qualifies as an author, students demonstrated a fairly nuanced understanding of this complex issue. One student felt that this was a gray area, citing that even outside of GAI generated content, scholars build on each other's work to such an extent that the originality of any work can be questioned. In case study responses, all 103 students agreed that using GAI without permission is dishonest. Respondents noted that students, instructors, and university administration each have a responsibility to ensure that GAI is not misused. And like peers elsewhere, they welcome institutional policy and guidance on using GAI.

Generative AI (GAI) can be leveraged to good effect in engineering, technology, and computing (ETC) classrooms but with pre-conditions; among these include teacher preparedness. One dimension of teacher preparedness is knowing how GAI is used in classrooms like their own. This paper responds to that need. The paper surveys the literature to answer three questions: in formal instructional interventions, what were the tasks for which GAI was used; what methods did scholars use to evaluate the GAI instructional set; and what were some of the identified challenges and opportunities regarding the use of GAI in ETC classrooms? Taking account of quality and recency of scholarship, and the need to focus on ETC practices, a rapid review of IEEE FIE conference proceedings for 2023 and 2024 was undertaken. Of the 1,181 papers published in both conference proceedings, and after a two-stage screening process, 20 papers were selected for synthesis. Of the 20 papers examined in this review, we found seven patterns of research design. The most common was the learning and teaching intervention followed by gathering student feedback. While we expect that a bigger corpus will yield additional research designs, these seven provide a start for defining a typology of research design investigating student use of GAI. The review revealed eight distinct categories of use ranging from software development, the most common, with over 26 distinct tasks, to data classification and data analysis, each with two distinct tasks. GAI was found to be used for many low-cognitive tasks like generating bibliographies to cognitivelydemanding uses like design and modeling. The studies reported recurring concerns about using GAI, like perpetuating bias, and hallucinations. One striking empirical finding is that previous ways of accomplishing ETC tasks, like using MATLAB and hand analysis, may be quicker and easier than using GAI. The scholarship advocates for broad and deep GAI literacy programs, addressing GAI abilities, limitations, ethics of use, and prompt design. Moreover, instructors are encouraged to be GAI literate themselves. This know-how is central to designing meaningful GAI-based tasks which seek to prepare students for an increasingly AI-infused world and workplace.

Contribution: This research, WIP paper describes a domain-specific, literature review on the use of Digital Twins (DTs) in undergraduate, Control Engineering Education (CEE). Such a body of work is currently absent from the scientific literature. The existing DT reviews were quantitatively assessed, and then the CEE applications, implementation possibilities, emerging opportunities, challenges, and future directions identified. Background: DTs are cutting-edge technologies, touted for their wide range of applications and services. As a result, domain-focused, DT reviews are often found in the scientific literature, to address the individual needs and challenges of various application domains. However, there is a need for such reviews in education and even CEE, which is itself tightly coupled to DT technology. Research Questions: 1. To what extent is education represented in recent, domain-specific, literature reviews on DTs as compared to other DT application domains? 2. What opportunities exist for the incorporation of DTs and their services into undergraduate, CEE and how can the challenges associated with their integration be addressed? Methodology: The methodology involved conducting a systematic literature review of peer-reviewed, domain-specific reviews published between 2019 and 2024 and then quantitatively as-sessing the representation of education and other application domains. Subsequently, peer reviewed, primary sources were syn-thesized into an integrative review on available DT services, CEE applications, emerging opportunities and challenges associated with DT implementations. Findings: This work has established that 1) there is a lack of application-focused, DT reviews in the field of education and its subdomain CEE 2) manufacturing and energy systems have the greatest prominence in existing DT reviews. 3) opportunities exist in learning theory-based DT design, novel enabling technologies, interacting DTs and sustainable education, and 4) challenges include the management of curricula, data, finances, change, and ethical issues, and 5) future works should develop frameworks that capitalise on the above opportunities.

Free-to-use generative AI (GAI) threatens assessment integrity. The scholarship establishes that GAI can produce passing to highly sophisticated responses to a range of assessment items. And detecting AI-generated output is fraught. Human detection is spotty and there are no proven software solutions at the time of writing. Even where there are promising detection solutions, these are likely to become obsolete as GAI evolves. The challenge for instructors is reliably and consistently establishing authorship of student submissions. There are two main perspectives on academic cheating. Proactive approaches appeal to students' honor and precede submission. Whereas punitive strategies are employed after submission, intending to detect and punish dishonesty. This paper focuses on post-assessment regimes where academic integrity is checked after submission. This work-in-progress, research-to-practice paper collates post-assessment strategies from the literature for detecting unsanctioned use of GAI. Two sets of scholarship were synthesized to answer one question: What post-assessment strategies are there for detecting unsanctioned use of GAI? The first set of literature focusses on post-assessment strategies in general. The second set addresses post-assessment strategies given the proliferation of free-to-use GAI. The Education Resources Information Center (ERIC) database was searched for general post-assessment strategies, yielding five tools. This search of general, non-discipline specific scholarship returns instructors to practices that have been tried and tested before the advent of GAI. It is possible that they can form part of a larger strategy of reducing GAI misuse; but this requires further study. The second set of tools, extracted from scholarship published by IEEE Xplore, totals six. These are post-assessment tools specific to detecting GAI-generated text. The study reveals that as scholars discuss the nature of GAI misuse, there is need to problematize definitions of that misuse. There are several tools available to instructors for detecting dishonest content; no single tool is infallible. Instructors need to think not in terms of a one-off detection solution but a regimen of post-assessment checks to compensate for this fallibility of tools. And there is the need to test tools that work best given teaching and learning contexts.

The estimation of temporal occupancy statistics is a common monitoring output for spectrum management. In emerging dynamic spectrum access (DSA) networks as well as for recent interference limit policies such as harm claim threshold mechanisms, this is even more crucial for compliance enforcement since operational parameters additionally can include temporal constraints. In this paper, random temporal sampling is explored, for probabilistic characterization of temporal channel occupancy. The precision and bias performance due to various random temporal measurement plan designs are examined, both analytically as well as through experiments implemented using Software Defined Radio technology. A framework for performance analysis of random temporal sampling for compliance enforcement is presented, and a lower bound on sensing performance in terms of estimator precision is derived for spectrum occupancy modeled as an alternating renewal process. Using random sampling, estimator variance is seen to approach the theoretical lower bound using larger sample sizes, or through sparser sampling. Results further suggest that the detection error impacts the performance of random temporal sampling for average temporal occupancy estimation. The work further motivates the use of probabilistic characterization of spectrum occupancy for compliance enforcement, given the non-deterministic behavior of dynamic spectrum access mechanisms in emerging wireless network deployment scenarios.

One challenge faced in the engineering education, is the need to imbue both technical competence and critical thinking skills within the bounds of academic program delivery. While technical competence can be built and assessed using structured and/or quantitative exercises, critical thinking is a skill that is both difficult to cultivate and to assess. Critical thinking involves the possession of both an expert mental model as well as the ability to leverage this model in various tasks. In engineering education, these tasks include making and justifying design choices, system optimization, and predicting system performance. Prior work, by one of the authors, explored the role of graphic organizers in the development of student's mental models. This paper describes action research underway, to explore the use of the argument map, as a structured means of leveraging mental models to promote critical thinking. In this paper, interactive small-group tasks based on argument maps are presented, and outputs generated by the initial cohort of undergraduate senior learners on these tasks are examined for evidence of critical thinking. These items form the basis of a longer-term longitudinal study in which the most effective means of deploying argument maps for promoting critical thinking will be examined.

Densely-deployed embedded sensor networks are susceptible to constraints associated with contention across a shared transport medium. To improve channel reliability, as well as average power consumption across the system, densely-deployed embedded sensor networks often leverage node-based neighbourhood data aggregation strategies. The tradeoff is that individual sensor nodes will have increased memory capacity and access requirements; where access requirements are determined by the memory transport bandwidth, the nature and frequency of the memory accesses, and the latencies associated with the memory storage mechanism. Individual sensor nodes consume power both directly based on the number/nature of memory operations, and indirectly through leakage current through latent circuitry. This paper considers the impact of different memory archetypes on performance of aggregation-related algorithms by individual nodes - specifically the scalability of number of required bus transactions and memory-related latencies with data-set size. The archetypes under consideration were: linear-addressing (RAM), content-based addressing (ternary CAM), and multi-dimensional addressing (Parks'). VHDL-specified MicroBlaze-based nodes, a 32 bit data-bus, and archetypical memories were implemented on a Virtex-5 development board. Operations central to aggregation algorithms (min, sum, count) were run using each type of memory on data-sets of 8 different sizes between 8 and 1024 data-points. Results suggest that appropriate selection of local-node memory architecture, can offer performance benefits in densely deployed sensor networks.

We propose a novel channel selection method for transmitting information that takes into account the amount of communication data and control data generated by the secondary users within a cognitive radio network. In this paper, multiple primary channels are characterized according to a channel occupancy ratio and a state transition ratio (STR), from which the secondary user selects the channel most suitable for transmitting the control information. By obtaining these ratios, the secondary user can estimate the size of a spectral white space for each channel belonging to the primary user. The proposed method is evaluated through computer simulation results and we can confirm the long-term white space can be remained for data transmission.