Search for: All records

Abstract We provide a novel characterization of augmented balancing weights, also known as automatic debiased machine learning. These popular doubly robust estimators combine outcome modelling with balancing weights—weights that achieve covariate balance directly instead of estimating and inverting the propensity score. When the outcome and weighting models are both linear in some (possibly infinite) basis, we show that the augmented estimator is equivalent to a single linear model with coefficients that combine those of the original outcome model with those from unpenalized ordinary least-squares (OLS). Under certain choices of regularization parameters, the augmented estimator in fact collapses to the OLS estimator alone. We then extend these results to specific outcome and weighting models. We first show that the augmented estimator that uses (kernel) ridge regression for both outcome and weighting models is equivalent to a single, undersmoothed (kernel) ridge regression—implying a novel analysis of undersmoothing. When the weighting model is instead lasso-penalized, we demonstrate a familiar ‘double selection’ property. Our framework opens the black box on this increasingly popular class of estimators, bridges the gap between existing results on the semiparametric efficiency of undersmoothed and doubly robust estimators, and provides new insights into the performance of augmented balancing weights. 

Introduction: The emergence and widespread adoption of generative AI (GenAI) chatbots such as ChatGPT, and programming assistants such as GitHub Copilot, have radically redefined the landscape of programming education. This calls for replication of studies and reexamination of findings from pre-GenAI CS contexts to understand the impact on students. Objectives: Achievement Goals are well studied in computing education and can be predictive of student interest and exam performance. The objective in this study is to compare findings from prior achievement goal studies in CS1 courses with new CS1 courses that emphasize the use of human-GenAI collaborative coding. Methods: In a CS1 course that integrates GenAI, we use linear regression to explore the relationship between achievement goals and prior experience on student interest, exam performance, and perceptions of GenAI. Results: As with prior findings in traditional CS1 classes, Mastery goals are correlated with interest in computing. Contradicting prior CS1 findings, normative goals are correlated with exam scores. Normative and mastery goals correlate with students’ perceptions of learning with GenAI. Mastery goals weakly correlate with reading and testing code output from GenAI. 

Generative AI (GenAI) is advancing rapidly, and the literature in computing education is expanding almost as quickly. Initial responses to GenAI tools were mixed between panic and utopian optimism. Many were fast to point out the opportunities and challenges of GenAI. Researchers reported that these new tools are capable of solving most introductory programming tasks and are causing disruptions throughout the curriculum. These tools can write and explain code, enhance error messages, create resources for instructors, and even provide feedback and help for students like a traditional teaching assistant. In 2024, new research started to emerge on the effects of GenAI usage in the computing classroom. These new data involve the use of GenAI to support classroom instruction at scale and to teach students how to code with GenAI. In support of the former, a new class of tools is emerging that can provide personalized feedback to students on their programming assignments or teach both programming and prompting skills at the same time. With the literature expanding so rapidly, this report aims to summarize and explain what is happening on the ground in computing classrooms. We provide a systematic literature review; a survey of educators and industry professionals; and interviews with educators using GenAI in their courses, educators studying GenAI, and researchers who create GenAI tools to support computing education. The triangulation of these methods and data sources expands the understanding of GenAI usage and perceptions at this critical moment for our community. 

Distributional checklists of the extant, described species of five superfamilies of Hymenoptera of Canada, Alaska and Greenland are presented. In total, 296 species in 79 genera in 12 families are recorded: 55 species of Ceraphronoidea, classified in 10 genera in 2 families, 205 species of Cynipoidea in 58 genera in 5 families, 30 species of Evanioidea in 5 genera in 3 families of Evanioidea, 2 species of Stephanoidea in 2 genera in 1 family and 4 species of Trigonalyoidea in 4 genera in 1 family. Of the reported species, 281 (in 79 genera in 12 families) are listed from Canada, 31 (in 16 genera in 6 families) from Alaska, and 7 (in 5 genera in 2 families) from Greenland. The list includes 8 new generic records for Canada (1 Ceraphronoidea, 6 Cynipoidea and 1 Evanioidea) and 43 new Canadian species records (13 Ceraphronoidea, 28 Cynipoidea and 2 Evanioidea). For each species in Canada, distribution is tabulated by province or territory, except the province of Newfoundland and Labrador is divided into the island of Newfoundland and the region of Labrador. These checklists are compared with previous Nearctic and Palaearctic surveys, checklists and catalogues.Kleidotoma minimaProvancher, 1883 (Figitidae) is moved from this genus toHexacolaFörster, 1869 to formH. minimum(Provancher, 1883),comb. nov.Amblynotus slossonaeCrawford, 1917 (Figitidae) is moved fromMelanipsWalker, 1835 toAmphithectusHartig, 1840 formingA. slossonae(Crawford, 1917),comb. nov. 

We report the growth of AlBN/β‐Nb₂N nitride epitaxial heterostructures in which the AlBN is ferroelectric, and β‐Nb₂N with metallic resistivity ≈40 μ at 300 K becomes superconducting belowT_C ≈ 0.5 K. Using nitrogen plasma molecular beam epitaxy, we grow hexagonal β‐Nb₂N films on c‐plane Al₂O₃substrates, followed by wurtzite AlBN. The AlBN is in epitaxial registry and rotationally aligned with the β‐Nb₂N, and the hexagonal lattices of both nitride layers make angles of 30° with the hexagonal lattice of the Al₂O₃substrate. The B composition of the AlBN layer is varied from 0 to 14.7%. It is found to depend weakly on the B flux, but increases strongly with decreasing growth temperature, indicating a reaction rate‐controlled growth. The increase in B content causes a non‐monotonic change in the a‐lattice constant and a monotonic decrease in the c‐lattice constant of AlBN. Sharp, abrupt epitaxial AlBN/β‐Nb₂N/Al₂O₃heterojunction interfaces and close symmetry matching are observed by transmission electron microscopy. The observation of ferroelectricity and superconductivity in epitaxial nitride heterostructures opens avenues for novel electronic and quantum devices. 

The ability of students to “Explain in Plain English” (EiPE) the purpose of code is a critical skill for students in introductory programming courses to develop. EiPE questions serve as both a mechanism for students to develop and demonstrate code comprehension skills. However, evaluating this skill has been challenging as manual grading is time consuming and not easily automated. The process of constructing a prompt for the purposes of code generation for a Large Language Model, such OpenAI’s GPT-4, bears a striking resemblance to constructing EiPE responses. In this paper, we explore the potential of using test cases run on code generated by GPT-4 from students’ EiPE responses as a grading mechanism for EiPE questions. We applied this proposed grading method to a corpus of EiPE responses collected from past exams, then measured agreement between the results of this grading method and human graders. Overall, we find moderate agreement between the human raters and the results of the unit tests run on the generated code. This appears to be attributable to GPT-4’s code generation being more lenient than human graders on low-level descriptions of code 

Abstract Most prior research characterizes information-seeking behaviors as serving utilitarian purposes, such as whether the obtained information can help solve practical problems. However, information-seeking behaviors are sensitive to different contexts (i.e., threat vs. curiosity), despite having equivalent utility. Furthermore, these search behaviors can be modulated by individuals' life history and personality traits. Yet the emphasis on utilitarian utility has precluded the development of a unified model, which explains when and how individuals actively seek information. To account for this variability and flexibility, we propose a unified information-seeking framework that examines information-seeking through the lens of motivation. This unified model accounts for integration across individuals' internal goal states and the salient features of the environment to influence information-seeking behavior. We propose that information-seeking is determined by motivation for information, invigorated either by instrumental utility or hedonic utility, wherein one's personal or environmental context moderates this relationship. Furthermore, we speculate that the final common denominator in guiding information-seeking is the engagement of different neuromodulatory circuits centered on dopaminergic and noradrenergic tone. Our framework provides a unified framework for information-seeking behaviors and generates several testable predictions for future studies. 

During the 2022 New Mexico monsoon season, we deployed two X‐ray scintillation detectors, coupled with a 180 MHz data acquisition system to detect X‐rays from natural lightning at the Langmuir Lab mountain‐top facility, located at 3.3 km above mean sea level. Data acquisition was triggered by an electric field antenna calibrated to pick up lightning within a few km of the X‐ray detectors. We report the energies of over 240 individual photons, ranging between 13 keV and 3.8 MeV, as registered by the LaBr3(Ce) scintillation detector. These detections were associated with four lightning flashes. Particularly, four‐stepped leaders and seven dart leaders produced energetic radiation. The reported photon energies allowed us to confirm that the X‐ray energy distribution of natural stepped and dart leaders follows a power‐law distribution with an exponent ranging between 1.09 and 1.96, with stepped leaders having a harder spectrum. Characterization of the associated leaders and return strokes was done with four different electric field sensing antennas, which can measure a wide range of time scales, from the static storm field to the fast change associated with dart leaders. 

We present analysis on two X-ray bright points observed over several hours during the recent solar minimum (2020 February 21 and 2020 September 12–13) with the Nuclear Spectroscopic Telescope Array (NuSTAR), a sensitive hard X-ray imaging spectrometer. This is so far the most detailed study of bright points in hard X-rays, emission which can be used to search for faint hot and/or non-thermal sources. We investigate the bright points’ time evolution with NuSTAR, and in extreme ultraviolet (EUV) and soft X-rays with Solar Dynamic Observatory/Atmospheric Imaging Assembly (SDO/AIA) and Hinode/X-Ray Telescope. The variability in the X-ray and EUV time profiles is generally not well matched, with NuSTAR detecting spikes that do not appear in EUV. We find that, for the 2020 February bright point, the increased X-ray emission during these spikes is due to material heated to ∼ 4.2–4.4 MK (found from fitting the X-ray spectrum). The 2020 September bright point also shows spikes in the NuSTAR data with no corresponding EUV signature seen by SDO/AIA, though in this case, it was due to an increase in emission measure of material at ∼ 2.6 MK and not a significant temperature change. So, in both cases, the discrepancy is likely due to the different temperature sensitivity of the instruments, with the X-ray variability difficult to detect in EUV due to cooler ambient bright point emission dominating. No non-thermal emission is detected, so we determine upper limits finding that only a steep non-thermal component between 3 and 4 keV could provide the required heating whilst being consistent with a null detection in NuSTAR.