Search for: All records

The Defining Issues Test 2 (DIT-2) and Engineering Ethical Reasoning Instrument (EERI) are designed to measure ethical reasoning of general (DIT-2) and engineering-student (EERI) populations. These tools—and the DIT-2 especially—have gained wide usage for assessing the ethical reasoning of undergraduate students. This paper reports on a research study in which the ethical reasoning of first-year undergraduate engineering students at multiple universities was assessed with both of these tools. In addition to these two instruments, students were also asked to create personal concept maps of the phrase “ethical decision-making.” It was hypothesized that students whose instrument scores reflected more postconventional levels of moral development and more sophisticated ethical reasoning skills would likewise have richer, more detailed concept maps of ethical decision-making, reflecting their deeper levels of understanding of this topic and the complex of related concepts. In fact, there was no significant correlation between the instrument scores and concept map scoring, suggesting that the way first-year students conceptualize ethical decision making does not predict the way they behave when performing scenario-based ethical reasoning (perhaps more situated). This disparity indicates a need to more precisely quantify engineering ethical reasoning and decision making, if we wish to inform assessment outcomes using the results of such quantitative analyses. 

In this paper, we present improvements to the pointing accuracy of the South Pole Telescope (SPT) using machine learning. The ability of the SPT to point accurately at the sky is limited by its structural imperfections, which are impacted by the extreme weather at the South Pole. Pointing accuracy is particularly important during SPT participation in observing campaigns with the Event Horizon Telescope (EHT), which requires stricter accuracy than typical observations with the SPT. We compile a training dataset of historical observations of astronomical sources made with the SPT-3G and EHT receivers on the SPT. We train two XGBoost models to learn a mapping from current weather conditions to two telescope drive control arguments — one which corrects for errors in azimuth and the other for errors in elevation. Our trained models achieve root mean squared errors on withheld test data of 2[Formula: see text]14 in cross-elevation and 3[Formula: see text]57 in elevation, well below our goal of 5[Formula: see text] along each axis. We deploy our models on the telescope control system and perform further in situ test observations during the EHT observing campaign in April 2024. Our models result in significantly improved pointing accuracy: for sources within the range of input variables where the models are best trained, average combined pointing error improved 33%, from 15[Formula: see text]9 to 10[Formula: see text]6. These improvements, while significant, fall shy of our ultimate goal, but they serve as a proof of concept for the development of future models. Planned upgrades to the EHT receiver on the SPT will necessitate even stricter pointing accuracy which will be achievable with our methods. 

Ethics education has been recognized as increasingly important to engineering over the past two decades, although disagreement exists concerning how ethics can and should be taught in the classroom. With active learning strategies becoming a preferred method of instruction, a collaboration of authors from four universities (University of Pittsburgh, University of Connecticut, Rowan University and New Jersey Institute of Technology) are investigating how game-based or playful learning with strongly situated components can influence first-year engineering students’ ethical knowledge, awareness, and decision making. This paper offers an overview and results of the progress to date of this three year, NSF Improving Undergraduate STEM Education (IUSE) grant that aims to (1) characterize the ethical awareness and decision making of first-year engineering students, (2) develop game-based learning interventions focused on ethical decision making, and (3) determine how (and why) game-based approaches affect students’ ethical awareness in engineering and the advantages of such approaches over non game-based approaches. Now in its second year, the authors have conducted a preliminary analysis of first-year students' ethical knowledge and organization via a concept mapping approach and have measured students' ethical reasoning using the Defining Issues Test 2 (DIT2) and Engineering Ethics Reasoning Instrument (EERI). Further, the authors have developed a suite of ethics-driven games that have been implemented across three of the universities, engaging over 400 first-year engineering students. Evaluation data has also been gathered for further game development and to assess initial student engagement and learning. Year 1 has provided insight into where first-year engineering students “are at” in terms of ethical knowledge and reasoning when they come to college, and how game-based instruction can be effective in the development of these students into moral agents who understand the consequences of their decisions. Further results from this investigation will provide the engineering education community with a set of impactful and research-based playful learning pedagogy and assessment that will help students confront social and ethical dilemmas in their professional lives. 

We propose a novel policy gradient method for multi-agent reinforcement learning, which leverages two different variance-reduction techniques and does not require large batches over iterations. Specifically, we propose a momentum-based decentralized policy gradient tracking (MDPGT) where a new momentum-based variance reduction technique is used to approximate the local policy gradient surrogate with importance sampling, and an intermediate parameter is adopted to track two consecutive policy gradient surrogates. MDPGT provably achieves the best available sample complexity of O(N -1 e -3) for converging to an e-stationary point of the global average of N local performance functions (possibly nonconcave). This outperforms the state-of-the-art sample complexity in decentralized model-free reinforcement learning and when initialized with a single trajectory, the sample complexity matches those obtained by the existing decentralized policy gradient methods. We further validate the theoretical claim for the Gaussian policy function. When the required error tolerance e is small enough, MDPGT leads to a linear speed up, which has been previously established in decentralized stochastic optimization, but not for reinforcement learning. Lastly, we provide empirical results on a multi-agent reinforcement learning benchmark environment to support our theoretical findings.