Program: Public Policy and Administration PhD

Committee Chair: Chris Birdsall

Committee: Chris Birdsall, Stephanie Witt, Jaclyn Kettler, Heather Williams

Abstract: This dissertation applies the Narrative Policy Framework to examine narrative strategies within the school choice legislative debate, specifically focusing on the devil-angel shift. The devil-angel shift is historically understood as a function of coalition status, with Shanahan et. al. (2013) hypothesizing that policy actors deploy this narrative device differently depending on whether they believe they are winning or losing the policy debate. In opposition to this hypothesis, Gottlieb et al. (2018) proposed an alternative hypothesis suggesting that a policy actor’s position, as opposed to their belief of winning or losing the policy debate, more accurately predicts the use of the devil-angel shift.

Using a quantitative content analysis of seven hundred nine legislative narratives from Idaho and Nevada education committee meetings, this study examines whether the devil-angel shift is present in school choice policy discussions and whether its use is associated with winning status, losing status, or policy position. Narrative elements were coded using an established modified codebook to identify heroes, villains, victims, and policy stances.

The findings indicate that the devil-angel shift is present in school choice policy debates, although hero and victim narratives appeared far more frequently than villain narratives. The results provide limited support for the argument that perceptions of winning or losing drive narrative strategy and stronger support for Gottlieb et al.'s (2018) alternative hypothesis that policy position is a more important predictor of narrative construction. The comparison between Idaho and Nevada also suggests that institutional context influences narrative element use and policy framing. The findings suggest that policy position and institutional context play a more significant role in shaping legislative narratives than previously theorized.

View on site | Email this event

Program: Computing PhD

Committee Chair: Alejandro Flores

Committee: Alejandro Flores, Hans-Peter Marshall, James McNamara, John Schreck

Abstract: Winter hydrometeorological prediction remains challenging because of complex interactions among atmospheric forcing, precipitation phase, atmosphere-land interaction, model uncertainty, sparse observations, and computational cost. This dissertation investigates these challenges by progressing from physics-based numerical modeling to artificial intelligence approaches.

First, this dissertation evaluates the sensitivity of mountain snowpack simulations to atmospheric model configuration using WRF–Noah-MP experiments over the Central Idaho Rocky Mountains. Results show that cloud microphysics schemes and lateral boundary conditions substantially influence precipitation, snow water equivalent (SWE), snow depth, and melt timing, demonstrating how uncertainty in atmospheric forcing propagates into land surface predictions. Second, we develop a spatiotemporal deep learning emulator to reproduce Noah-MP SWE over the Colorado Rocky Mountains using ConvLSTM networks driven by WRF meteorological forcings, physiographic variables, and satellite-derived observations. The emulator captures spatial and seasonal SWE variability, outperforms a pixelwise LSTM, and provides substantially faster predictions than the parent physics-based model. Third, we assess the ability of five AI-based numerical weather prediction systems to support precipitation-type diagnosis using an evidential deep neural network and mPING observations. AI weather models maintain stronger skill for rain and snow than for freezing rain and ice pellets, with mixed-phase performance degrading rapidly at longer lead times as shallow thermodynamic structures are lost. Uncertainty estimates also reveal that increasing model confidence does not necessarily correspond to increasing forecast reliability.

Together, these studies demonstrate the complementary strengths and limitations of physics-based and AI-based approaches and highlight pathways toward more accurate, efficient, and uncertainty-aware prediction of winter hydrometeorological processes.

View on site | Email this event

Program: Computing Ph.D.

Committee Chair: Alejandro Flores

Committee: Alejandro Flores, Hans-Peter Marshall, James McNamara, John Schreck

Abstract: Winter hydrometeorological prediction remains challenging because of complex interactions among atmospheric forcing, precipitation phase, atmosphere-land interaction, model uncertainty, sparse observations, and computational cost. This dissertation investigates these challenges by progressing from physics-based numerical modeling to artificial intelligence approaches.

First, this dissertation evaluates the sensitivity of mountain snowpack simulations to atmospheric model configuration using WRF–Noah-MP experiments over the Central Idaho Rocky Mountains. Results show that cloud microphysics schemes and lateral boundary conditions substantially influence precipitation, snow water equivalent (SWE), snow depth, and melt timing, demonstrating how uncertainty in atmospheric forcing propagates into land surface predictions. Second, we develop a spatiotemporal deep learning emulator to reproduce Noah-MP SWE over the Colorado Rocky Mountains using ConvLSTM networks driven by WRF meteorological forcings, physiographic variables, and satellite-derived observations. The emulator captures spatial and seasonal SWE variability, outperforms a pixelwise LSTM, and provides substantially faster predictions than the parent physics-based model. Third, we assess the ability of five AI-based numerical weather prediction systems to support precipitation-type diagnosis using an evidential deep neural network and mPING observations. AI weather models maintain stronger skill for rain and snow than for freezing rain and ice pellets, with mixed-phase performance degrading rapidly at longer lead times as shallow thermodynamic structures are lost. Uncertainty estimates also reveal that increasing model confidence does not necessarily correspond to increasing forecast reliability.

Together, these studies demonstrate the complementary strengths and limitations of physics-based and AI-based approaches and highlight pathways toward more accurate, efficient, and uncertainty-aware prediction of winter hydrometeorological processes.

View on site | Email this event

Program: Biology MS

Committee Chair: Jodi Brandt

Committee: Jodi Brandt, Trevor Caughlin, Jen Cruz, Charles Hawkins

Abstract: Humans have dramatically changed seasonal variability across many ecosystems. Seasonal variability
may increase biodiversity by generating heterogeneous habitats, yet also reduce aquatic diversity by
intensifying water limitation and habitat contraction. Using nine years of seasonal remote-sensing data
paired with 1,718 stream surveys, we tested whether macroinvertebrate biodiversity in drylands
responds more strongly to water availability or seasonal variability. Seasonal variability had a larger
main effect on macroinvertebrate communities than water availability, and was comparable to
established environmental drivers, including pH and elevation. Water availability and seasonal variability
interacted: seasonal variability increased richness most where water was scarce and had little effect
where water was abundant. Taxa responded uniquely to water availability and seasonal variability,
indicating that distinct hydrologic regimes support different components of biodiversity. Recognizing
seasonal variability as a fundamental dimension of ecological state can improve how science, policy, and
management identify and protect the hydrologic regimes that sustain biodiversity.

View on site | Email this event

Program: Educational Technology EdD

Committee Chair: Ross Perkins

Committee: Ross Perkins, Andy Hung, Yu-hui Ching

Abstract: Beginning-level learners of Japanese as a Foreign Language (JFL) often have limited opportunities for systematic pronunciation and prosody instruction, scaffolded oral practice, and individualized feedback. This quantitative, quasi-experimental study examined adding prosody-focused shadowing plus automatic speech recognition (ASR) to regular instruction. The analytic sample included 50 students in four intact Japanese 101 classes at two public universities, with one experimental and one control class per site. Over approximately 10 weeks, experimental groups used ASR output for formative checking and revision during shadowing. Pretests, posttests, questionnaires, and weekly logs provided data. For each university, analyses of covariance adjusted for corresponding pretest scores and the Out-of-Class Japanese Use score. Listening and speaking anxiety and overall self-efficacy hypotheses were not supported. The adjusted passage-level listening advantage was significant only at University A and required caution due to unequal error variances. Experimental groups showed significant adjusted advantages in composite oral performance at both universities. The routine may support oral performance in beginning-level JFL instruction.

View on site | Email this event

Program: Electrical and Computer Engineering PhD

Committee Chair: David Estrada

Committee: David Estrada, Kurtis Cantley, Josh Eixenberger, Stephen Parke

Abstract: Additive manufacturing of functional nanomaterials offers new opportunities for the fabrication of electronic devices by reducing process complexity and material waste while enabling compatibility with conventional substrates. Digital-write approaches are particularly attractive because they enable the deposition of functional materials with high spatial resolution on a wide range of substrates and material formulations. This dissertation investigates additive manufacturing approaches for the processing, deposition, and functional integration of two-dimensional (2D) and nanostructured materials, including graphene, nanodiamond, laser-induced graphene (LIG), and hexagonal boron nitride (hBN), for hybrid electronic devices. (1) The first part of this work focuses on the formulation and aerosol jet printing of graphene-based inks for capacitive touch sensors and Joule heating applications. Graphene inks were developed and optimized for AJP, enabling the printing of high-resolution features with line widths of approximately 5 µm without post-processing through process parameter variation. This was followed by the printing of structures for transmission line method (TLM) measurements, achieving a low sheet resistance of approximately 110 O/?, as well as the fabrication of conductive interdigitated electrode (IDE) and serpentine structures for capacitive touch sensing and Joule heating applications, demonstrating the versatility of aerosol jet-printed graphene for multifunctional electronic applications. (2) The second objective investigates the aerosol jet printing of nanodiamond thin films and their frequency-dependent dielectric behavior using commercially available modified detonated nanodiamond inks. AJP-printed films were characterized over a range of frequencies to establish the relationship between printed film structure and dielectric response. This work provides insights into the electrical properties of AJP-printed nanodiamond and establishes a foundation for its integration into functional electronic structures. (3) The third objective explores the integration of carbon- and nitride-based materials through complementary additive manufacturing approaches. Laser-induced graphene (LIG) was employed to produce conductive structures, while hexagonal boron nitride (hBN) was deposited using extrusion printing to provide electrically insulating and dielectric functionality. These materials were integrated to demonstrate simple resistor-capacitor (RC) circuits, illustrating the ability to combine conductive and insulating 2D materials within additively manufactured electronic structures.
Collectively, the results demonstrate the potential of additive manufacturing to enable the scalable fabrication and functional integration of diverse nanomaterials for multifunctional electronic devices.

View on site | Email this event

Program: Computing Ph.D.

Committee Chair: Matt Williamson

Committee: Matt Williamson, Francesca Spezzano, Chris Birdsall

Abstract:

Biodiversity loss and species extinction remain urgent environmental challenges, making effective species-protection policy increasingly important in the United States. Yet the implementation and public understanding of such policy are shaped not only by scientific and legal evidence, but also by how endangered species issues are communicated and contested across media and political institutions. In the case of the U.S. Endangered Species Act (ESA), these public debates are shaped not only by scientific and legal information but also by narratives that define problems, assign responsibility, and advocate policy responses. These narratives are distributed across large collections of news and congressional text, making text analysis difficult with manual coding alone. This dissertation develops a computational framework for examining how ESA narratives are constructed, circulate between media and political institutions, and relate to policy outcomes. I pursue this goal through three objectives. Objective 1 develops and validates a prompt-engineered large language model (LLM) pipeline for extracting Narrative Policy Framework (NPF) elements from national news and Congressional Record text. Objective 2 uses the resulting extracted narratives to examine directional and reciprocal agenda-setting relationships between news coverage and congressional attention across multiple temporal lag windows and publication outlets. Objective 3 links media attention and narrative characteristics to the timing of formal ESA decisions using survival analysis of species listing and critical habitat actions. Together, the three objectives move from measurement to cross-institutional narrative agenda setting, to policy consequences, providing an integrated computational approach to the study of environmental policy discourse.

View on site | Email this event

Program: Computing PhD

Committee Chair: Francesca Spezzano

Committee: Francesca Spezzano, Nasir Eisty, Bogdan Dit

Abstract: Scientific software underpins modern discovery, from climate modeling to computational physics, yet its development often involves acknowledged shortcuts, unresolved assumptions, and temporary compromises. These admissions, known as Self-Admitted Technical Debt (SATD), pose unique risks where software correctness directly affects research validity and reproducibility. While SATD has been studied extensively in general-purpose open-source software (OSS), existing taxonomies, classifiers, and prioritization approaches were not designed for scientific contexts, leaving it unclear how SATD in scientific software should be identified, which debt receives attention, and how much of it persists. This dissertation investigates SATD in scientific software through three studies spanning identification, prioritization, and resolution. First, scientific debt was formally operationalized and validated with
domain practitioners as a domain-specific form of SATD, and a multi-artifact dataset of 115,524 labeled texts was constructed. A multi-task transformer classifier applied to 900,358 artifacts from 23 scientific software projects showed that SATD is most prevalent in pull requests and issues, and that a classifier limited to traditional SATD categories misses 74% of scientific debt instances. Because identification alone does not indicate which debt matters, the second study developed an embedding-based prioritization heuristic and traced SATD across linked artifacts in nine projects. Debt in code-proximate artifacts and debt expressed with negative sentiment received higher priority, and cross-artifact propagation was rare but marked higher priority debt. Only 37.8% of SATD comments were ever removed, and those that were persisted for a mean of roughly 690 days, both worse than OSS benchmarks. The third study examined what drives this persistence by comparing nine scientific and nine matched OSS projects using regression, survival models, and static analysis. Resolution was determined primarily by structural factors at introduction, such as file path type and maturity, rather than by author or sentiment. Documentation and scientific debt showed a compounding resolution disadvantage with age that was not observed in OSS, and SATD in scientific software was introduced into files with 2.25 times the SQALE debt ratio and 2.7 times the code smells of OSS files. Together, these findings demonstrate that SATD in scientific software differs qualitatively from that in general-purpose OSS: it is under-detected by conventional approaches, persistent, and prone to entrenchment. The datasets, models, and empirical evidence presented here give practitioners and researchers a foundation for domain-aware tracking, monitoring, and time-bounded escalation of SATD, improving the maintainability and reproducibility of scientific software.

View on site | Email this event

Program: Computing Ph.D.

Committee Chair: Francesca Spezzano

Committee: Francesca Spezzano, Nasir Eisty, Bogdan Dit

Abstract: Scientific software underpins modern discovery, from climate modeling to computational physics, yet its development often involves acknowledged shortcuts, unresolved assumptions, and temporary compromises. These admissions, known as Self-Admitted Technical Debt (SATD), pose unique risks where software correctness directly affects research validity and reproducibility. While SATD has been studied extensively in general-purpose open-source software (OSS), existing taxonomies, classifiers, and prioritization approaches were not designed for scientific contexts, leaving it unclear how SATD in scientific software should be identified, which debt
receives attention, and how much of it persists. This dissertation investigates SATD in scientific software through three studies spanning identification, prioritization, and resolution. First, scientific debt was formally operationalized and validated with
domain practitioners as a domain-specific form of SATD, and a multi-artifact dataset of 115,524 labeled texts was constructed. A multi-task transformer classifier applied to 900,358 artifacts from 23 scientific software projects showed that SATD is most
prevalent in pull requests and issues, and that a classifier limited to traditional SATD categories misses 74% of scientific debt instances. Because identification alone does not indicate which debt matters, the second study developed an embedding-based prioritization heuristic and traced SATD across linked artifacts in nine projects. Debt in code-proximate artifacts and debt expressed with negative sentiment received
higher priority, and cross-artifact propagation was rare but marked higher priority debt. Only 37.8% of SATD comments were ever removed, and those that were
persisted for a mean of roughly 690 days, both worse than OSS benchmarks. The third study examined what drives this persistence by comparing nine scientific and
nine matched OSS projects using regression, survival models, and static analysis. Resolution was determined primarily by structural factors at introduction, such as
file path type and maturity, rather than by author or sentiment. Documentation and scientific debt showed a compounding resolution disadvantage with age that was not observed in OSS, and SATD in scientific software was introduced into files with 2.25
times the SQALE debt ratio and 2.7 times the code smells of OSS files. Together, these findings demonstrate that SATD in scientific software differs qualitatively from that in general-purpose OSS: it is under-detected by conventional approaches, persistent, and prone to entrenchment. The datasets, models, and empirical evidence presented
here give practitioners and researchers a foundation for domain-aware tracking, monitoring, and time-bounded escalation of SATD, improving the maintainability and reproducibility of scientific software.

View on site | Email this event

Program: Materials Science and Engineering MS

Committee Chair: Elton Graugnard

Committee: Elton Graugnard, Steven Hues, David Estrada

Abstract:

The scaling of advanced microelectronic devices has created a demand for dielectric materials with improved material properties beyond those of conventional high-k dielectrics. Nitride materials, including aluminum nitride (Al4N3), gallium nitride (GaN), and scandium-alloyed aluminum nitride (Al0.88Sc0.12N), exhibit promising properties such as tunable bandgaps (3.2-6.3 eV), dielectric permittivity (ε = 12.8-17) pyroelectric coefficient (p = 15-20 μC/(m2K), piezoelectric responses (d33 = 24-28 pC/N), and high corrosion resistance. However, integrating these materials into increasingly scaled devices, particularly three-dimensional (3D) logic and memory architectures, presents significant processing challenges. Conventional deposition methods for improving film crystallinity, phase, purity, and defect density often require elevated temperatures or post-deposition annealing, which can exceed the thermal budget of back-end-of-line (BEOL) integration.

Atomic layer deposition (ALD) provides precise control over film thickness, uniformity, and composition but may require additional thermal treatment to achieve desired material properties. However, integrating controlled plasma-assisted atomic layer annealing (ALA) during PE-ALD steps may mitigate these challenges. During ALA, an Ar+ plasma promotes atomic rearrangement, densification, crystallization, and defect reduction through ionic-energy transfer at reduced deposition temperatures (<400 ℃). The Ar+ energy is maintained below the sputtering threshold to interact with and modify the growing film surface.

As part of this thesis, I developed an integrated plasma-assisted PE-ALD reactor for low-temperature ALA of nitride thin films. I used AlNₓ as a test vehicle to benchmark equipment and processes and to investigate ALA effects within individual PE-ALD cycles. Additionally, I integrated in-situ spectroscopic ellipsometry (SE) and residual gas analysis (RGA) to monitor film evolution, precursor chemistry, surface reactions, and byproducts in real time. The ALA approach with in-situ metrology enables investigation of plasma-assisted ALD advanced processing needed to support improved thin films process control within BEOL thermal budgets.

View on site | Email this event

Program: Chemistry MS

Committee Chair: Owen McDougal

Committee: Owen McDougal, Brian McClain, Henry Charlier

Abstract: Bean hydration is a time-intensive processing step that improves nutritional quality, speeds cooking time, and enhances texture development. Though necessary, this step creates a processing bottleneck that limits throughput and operational efficiency. The purpose of this research was to determine whether pulsed electric field (PEF) technology could accelerate the hydration of garbanzo, pinto, and black beans while maintaining bean quality. Beans were treated under low- and high-energy PEF conditions, 3 kV/cm, 8 kJ/kg and 3 kV/cm, 150 kJ/kg respectively, and compared with conventionally hydrated controls. The low-energy PEF treatment had minimal effects, while the high-energy PEF treatment reduced hydration times by approximately 45-90 min (27-33%) across the three bean varieties. The high-energy PEF-treated beans were further evaluated for color, texture, total solids, and physical integrity to assess quality. Quality remained consistent across control and PEF-treated beans with the exception of color, where more browning occurred in the PEF-treated pinto and garbanzo varieties. These findings demonstrate the potential for high-energy PEF treatment to accelerate dry bean hydration while maintaining important quality characteristics, thereby improving processing efficiency and throughput.

View on site | Email this event

Program: Computing PhD

Committee Chair: Steven Cutchin

Committee: Steven Cutchin, Jerry Alan Fails, Casey Kennington

Abstract: 3D Gaussian Splatting (3DGS) and Triangle Splatting represent a scene with millions of explicit primitives that are optimized from photographs and rendered in real time. While these methods produce high-quality images, they optimize geometry and appearance together, so a trained model does not show what each part contributes. To address this issue, this dissertation uses controlled changes to geometry and appearance to examine their roles during training, storage, and display. In 3DGS training, luminance is an effective signal for learning geometry that supports near-RGB reconstruction quality. After a common appearance refit, luminance-only geometry supports held-out reconstruction just 0.085 ± 0.037 dB below RGB-trained geometry on average, while chroma-only geometry falls several decibels below luminance-only geometry on the tested scenes. For Triangle Splatting, a post-training codec is proposed that derives its coding structure from decoded geometry. At the quality target set by a matched 3DGS reference codec, it reaches 18.5–30.1× compression on the eleven scenes where this target is bracketed. Although appearance dominates the raw model, geometry becomes the largest component of the compressed file. For high-resolution display, direct Mip-Splatting rendering provides a strong quality and cost baseline: it exceeds every tested per-frame super-resolution (SR) network on the 13-scene means of PSNR, SSIM, and LPIPS at substantially lower inference cost. Of the tested pipelines, only SR inside training (SRGS) exceeds it on PSNR, by 0.35 dB in the same latency tier. Refitting only the appearance of the Mip-Splatting model to SwinIR labels, with a fixed least-squares procedure, raises held-out PSNR on all nine Mip-NeRF 360 scenes. The gain averages 0.36 dB over the trained model drawn by the same rasterizer and 0.34 dB over direct rendering, although it depends on the representation, the labels, and the metric. Together, these findings identify practical priorities for the tested splatting pipelines: luminance can guide geometry learning, geometry warrants greater attention for further storage reduction, and appearance refinement can improve reconstruction without changing geometry.

View on site | Email this event

Program: Civil Engineering MS

Committee Chair: Bhaskar Chittoori

Committee: Bhaskar Chittoori, Nick Hudyma, Yang Lu

Abstract: Expansive clays damage pavements and lightweight structures through seasonal swelling and shrinkage driven by moisture change. Conventional stabilizers such as lime and cement work by altering the clay itself. This study evaluates calcium stearate, a hydrophobic compound, as an alternative that limits water entry rather than changing soil chemistry. It was compared with lime and cement on a high-plasticity clay from Marsing, Idaho. The study tests the hypothesis that calcium stearate reduces apparent swelling by slowing moisture ingress but does not reduce the clay’s intrinsic swell potential under confinement. The findings show that calcium stearate functions as a moisture barrier rather than a bulk stabilizer. They point to its use as a thin layer, possibly paired with lime, rather than as a soil-wide admixture.

View on site | Email this event

Program: Biomolecular Sciences PhD

Committee Chair: Juliette Tinker

Committee: Juliette Tinker, Ken Cornell, Javier Ochoa-Repáraz

Abstract: AB5 toxins are a structurally conserved family of bacterial proteins that contribute to host-pathogen interactions through an enzymatically active A-subunit and a pentameric B-subunit responsible for host-cell binding. Several well-characterized AB5 toxins, including cholera toxin (CT) from Vibrio cholerae, pertussis toxin (PT) from Bordetella pertussis, and heat-labile toxin (LT) from enterotoxigenic Escherichia coli (ETEC), function as major virulence factors by altering host-cell signaling and physiology. However, their structural and immunomodulatory properties have also made AB5 toxins attractive components for vaccine development. We hypothesized that AB5 toxins exert distinct effects on host-cell physiology during bacterial pathogenesis and that their structural and immunomodulatory properties can be adapted for vaccine development. This dissertation examines these contrasting roles by investigating the cellular signaling and functional activity of the Salmonella AB5 toxin ArtAB and expanding novel CT-derived vaccine platforms to prevent Staphylococcus aureus infection. Previous work from our laboratory purified ArtAB and its B-subunit, ArtB, from multidrug-resistant Salmonella Typhimurium definitive type 104 (DT104) and demonstrated distinct morphological effects in mammalian cells. To further characterize these effects, multi-omics and functional analyses were used to examine host-cell responses to ArtAB and ArtB. ArtAB produced minimal cytotoxicity but induced substantial host-cell remodeling and disrupted epithelial barrier integrity, supporting a role in modulating host physiology without overt cell death. In contrast, the mechanism of action of CT is well known, and it has been extensively studied for its adjuvant properties. Our laboratory uses a CT-derived CTA2/B chimeric vaccine platform in which the toxic CTA1 domain is replaced with an antigen of interest while retaining the CTA2 linking domain and non-toxic B-subunit (CTB). This platform was expanded and optimized for S. aureus vaccine development through recombinant chimera production, evaluation in a bovine mastitis vaccination and challenge model, and investigation of microneedle (MN)-based vaccine delivery for human applications. Platform optimization improved the production of selected recombinant vaccine constructs, while bovine vaccination elicited strong systemic antigen-specific serum IgG responses that were maintained through challenge. MN vaccination in mice produced limited antibody responses under the conditions tested but established feasibility for optimization of the delivery platform. Together, these studies demonstrate how AB5 toxins can function as pathogenic effectors while also providing adaptable platforms for vaccine development.

View on site | Email this event

Title: Integrating AI Coding Agents into Scientific Research

Presented by: Rory Barnes, Ph.D., Astronomy, University of Washington - Professor of Astronomy, Astrobiology, Data Science

Abstract:

New software tools have emerged that enable AI agents to quickly develop code that analyzes data and generates figures, but their mistakes can also harm a researcher's computer and lead to enhanced skepticism of AI-assisted results. These concerns could be mitigated if scientific software is developed in a secure environment with a system for monitoring changes in files and results. I explore these issues through the lenses of reproducibility standards, software engineering best practices, and artifact custody methods. I first identify a dozen ways in which AI agents change how scientific software and its results are generated, verified, and reproduced. I then propose a six-level scale that classifies AI-assisted numerical experiments by their provenance, reproducibility, openness, oversight, and falsifiability (PROOF). The scale begins with a locally self-consistent version-controlled code repository and culminates in a publicly accessible and curated archive of the files, change history, prompt transcripts, independent agent regeneration of the analysis, and multiple cryptographically verifiable attestations that the results have not been corrupted by AI agents. To help facilitate this safe and reproducible development of numerical experiments with AI agents, I created an open-source software package called vaibify that builds secure environments connected to a graphical interface where scientists can prompt command line agents, view the current status of the local and remote files, display figures, and be guided up the first three rungs of the PROOF Ladder. While the veracity of scientific results remains the domain of humans (for now), vaibify can at least ensure that the digital artifacts were generated in the correct order, local files match remote files, and the results are reproducible at the bit level over decade-long timescales.

Speaker Bio:

Rory Barnes is a professor of Astronomy, Astrobiology, and Data Science at the University of Washington in Seattle. His research focuses on developing computational models of planetary system evolution, with a focus on planetary habitability. He has also developed software packages that leverage machine learning to infer Bayesian uncertainties in high-dimensional problems. Lately, he has turned his attention toward exploiting AI coding agents to improve the quality of scientific software. Prof. Barnes began his career at the University of Arizona and received his Ph.D. in astronomy from the University of Washington, joining the faculty in 2013. Rory Barnes is a professor of Astronomy, Astrobiology, and Data Science at the University of Washington in Seattle. His research focuses on developing computational models of planetary system evolution, with a focus on planetary habitability. He has also developed software packages that leverage machine learning to infer Bayesian uncertainties in high-dimensional problems. Lately, he has turned his attention toward exploiting AI coding agents to improve the quality of scientific software. Prof. Barnes began his career at the University of Arizona and received his Ph.D. in astronomy from the University of Washington, joining the faculty in 2013.

View on site | Email this event

Program: Raptor Biology MS

Committee Chair: Stephen Novak

Committee: Stephen Novak, James Belthoff, James Smith

Abstract: Rapid global environmental change has caused drastic population declines, the extinction of species, and a reduction in biological diversity around the globe. One critical component of global biodiversity is the level of genetic diversity within populations, which influences the ability of populations to respond to changing conditions. This study offers unique insights into the genetics of a primarily migratory breeding population of Burrowing Owls (Athene cunicularia subsp. hypugaea) located in Southwestern Idaho, USA, through a genetic monitoring approach. The study included a total of 54 individuals sampled across five time periods (1997, 2002, 2007, 2013, and 2025). A 5,173 neutral Single Nucleotide Polymorphism (SNP) panel was assembled using Restriction site-Associated DNA sequencing (RADseq) methods to estimate levels of genetic diversity and genetic differentiation of this population, and assess whether genetic diversity has changed across 28 years. Multiple structure analyses (PERMANOVA, PCA, DAPC, STRUCTURE, and AMOVA) consistently showed an absence of temporally-based genetic structure. AMOVA detected no measurable genetic differentiation among sampling years, with essentially all genetic variation occurring “within individuals.” Pairwise Weir–Cockerham FST values remained close to zero (-0.0027 to 0.0015), supporting the conclusion that no temporal genetic structure has occurred. FIS values near zero, and slightly negative (ranging from -0.0128 to -0.0026), indicated that levels of inbreeding did not change over time. Other genetic diversity parameters (H?, H?, nucleotide diversity (p)) remained stable across all sampling years. These results indicate that individual owls sampled across the five time intervals represent a subsample of a larger, panmictic, and genetically diverse Snake River Plains population of Burrowing Owls, suggesting high connectivity and gene flow among individuals in this region. The study’s findings provide evidence that this breeding population has maintained stable genome-wide neutral genetic diversity across nearly three decades despite ongoing habitat change. This research also establishes a valuable genetic baseline for future genetic monitoring and conservation of this population, and other populations of Burrowing Owls.

View on site | Email this event

Program: Materials Science and Engineering PhD

Committee Chair: Zhangxian Deng

Committee: Zhangxian Deng, Eric Jankowski, Dave Estrada, Bob Fox

Abstract: Magnetostrictive terbium–iron–dysprosium alloys, commonly known as Terfenol-D, deform in response to magnetic fields and exhibit changes in magnetization under mechanical stress. This instantaneous magneto-mechanical coupling enables the generation and detection of various acoustic waveforms for structural health monitoring. However, the use of Terfenol-D in acoustic transducers is limited by integration challenges and its susceptibility to oxidation during manufacturing.

The objective of this research is to develop a comprehensive additive manufacturing process for magnetostrictive Terfenol-D while minimizing material oxidization, and to gain an understanding of how the printed Terfenol-D oxidizes in neutron radiation. First, an ink composed of Terfenol-D nanoparticles with a predominant particle size of approximately 155 nm was synthesized via high-energy ball milling and printed onto metal substrates using a commercial material extrusion (MEX) system. The printed samples exhibited approximately 3 wt.% oxidation when stored in an inert environment, indicating that oxidation was negligible before sintering. Second, the as-printed samples were thermally sintered in a tube furnace under a forming-gas atmosphere, followed by neutron irradiation for up to 6 days in the Ohio State University Research Reactor. Significant oxidation occurred during sintering, with subsequent irradiation promoting further oxide formation and amorphization. Finally, oxidation during thermal sintering was investigated using vacuum, glovebox, and turbo vacuum furnaces, as well as electric field-assisted sintering (EFAS). EFAS achieved strong film–substrate bonding, while turbo vacuum sintering showed promise in reducing oxide formation.

This work demonstrates a MEX-based route for fabricating magnetostrictive thin films, expands the additive manufacturing material library, and provides insights into oxidation behavior and strategies for mitigating oxidation during Terfenol-D processing.

View on site | Email this event

Program: Curriculum and Instruction EdD

Committee Chair: Jeremy W. Ford

Committee: Jeremy W. Ford, Katherine Wright, Jim Fredricksen

Abstract: This study was prompted by a real classroom dilemma: Why do secondary students who have all achieved proficiency on large-scale testing produce significantly different constructed responses when reading the same fictional text? A quantitative study was launched to examine the construct validity of the Higher Order Thinking-Reading Scale (HOT-RS), a theoretically grounded rubric designed to measure higher-order thinking in secondary readers’ constructed responses to reading a fictional text. Selected-response reading assessments capture accuracy and recall but not the depth of a reader’s thinking, and no validated instrument existed to measure higher-order thinking directly from student writing.
The HOT-RS was grounded in Afflerbach et al.’s (2015) framework for higher-order thinking in reading and developed following DeVellis’s (2017) scale development procedures. Participants were 49 secondary readers in grades 9 through 12, enrolled in an elective reading course at an urban public magnet high school in the northern Mountain West.
Over four weeks of ordinary instruction, students produced written journal responses with minimal prompting to self-selected fiction. Three trained raters scored these responses using the HOT-RS, achieving good interrater reliability (average-measures intraclass correlation = .895 at the student level, .821 at the journal level). HOT-RS scores were then correlated with scores on the Reading Maturity Survey (RMS), a validated 60-item self-report measure of reading maturity, to gather initial evidence of construct validity.
HOT-RS scores correlated positively and significantly with total RMS scores, r(47) = .35, p = .013, a medium effect. Of the six RMS subcomponents, HOT-RS scores correlated significantly with Reading Ability, r(47) = .51; Transformational Reading, r(47) = .37; and Reading Attitudes and Interests, r(47) = .33. Correlations with Reading Purposes, Sophistication of Reading Materials, and Higher-Order Literacy were positive but not significant.
This pattern offered both convergent and discriminant validity evidence that HOT-RS scores related most strongly to what a reader demonstrates and least strongly to what a reader self-reports. Convergent and discriminant evidence together provide initial construct validity evidence for the HOT-RS and suggest that higher-order thinking in reading, as this instrument measures it, is meaningfully related to, but distinct from, reading maturity.
The HOT-RS offers teachers a tool to measure higher-order thinking in reading as a means to understanding the thinking skills that readers possess, as well as those they lack. For researchers studying reading development, it offers an insight into the role that higher-order thinking plays in moving readers from proficiency towards that of expert reading.

View on site | Email this event