36214 research outputs found
Sort by
Cassia Blossoms
No full textHolograph (photocopy).Words by Li Ch'ing-Chao, translated from Chinese, also printed as text preceding score.At end: March 1988, Seattle, Washington.Premiered at the Bennington Chamber Music Conference and Composer Forum of the East, Aug. 10, 1988.Duration: 9:00.Contact [email protected] for more informatio
Position: Knowing Isn't Understanding: Re-grounding Generative Proactivity with Epistemic and Behavioral Insight
Full text linkGenerative AI agents equate understanding with resolving explicit queries, an assumption that confines interaction to what users can articulate. This assumption breaks down when users themselves lack awareness of what is missing, risky, or worth considering. In such conditions, proactivity is not merely an efficiency enhancement, but an epistemic necessity. We refer to this condition as epistemic incompleteness, where progress depends on engaging with unknown \textit{unknowns} for effective partnership.
Existing approaches to proactivity remain narrowly anticipatory, extrapolating from past behavior and presuming that goals are already well defined, thereby failing to support users meaningfully. However, surfacing possibilities beyond a user’s current awareness is not inherently beneficial. Unconstrained proactive interventions can misdirect attention, overwhelm users, or introduce harm. Proactive agents, therefore, require behavioral grounding: principled constraints on when, how, and to what extent an agent should intervene. We advance the position that generative proactivity must be grounded both epistemically and behaviorally. Drawing on the philosophy of ignorance and research on proactive behavior, we argue that these theories offer critical guidance for designing agents that can engage responsibly and foster meaningful partnerships
Native bee richness increases with wildfire burn severity in ponderosa pine forests.
No full textWildfires are increasing in frequency and severity in dry forests across western North America and have direct effects on forest structure and ecosystem services. One important service to monitor is pollination, which promotes plant-pollinator biodiversity and is critical for post-fire vegetative recovery. Because pollination services can vary by species and across spatial and temporal scales, understanding the effect of fire on pollinator populations informs conservation management and ecosystem restoration. Due to the known positive effects of fire on bees, including increased floral resources, nesting habitat, and light availability, we hypothesized that pollinator richness would increase with burn severity (measured with the Composite Burn Index) and decrease with burn age. We quantified native bee richness in the Okanogan-Wenatchee National Forest in Washington, USA at nine forested plots that burned in 2021, 2018, and 2015, or were unburned since 1968. Data were collected in 2021, 2022, and 2023 biweekly from April to August using blue vane and pan traps. Our findings suggest that native bee genus richness was driven by the interaction between burn severity and burn age. Study areas that had recently burned at higher severities had greater bee genus richness. Furthermore, the proportion of above ground nesting bees in landscapes 1-8 years post-fire was greatest at sites that burned with moderate fire severity and had more available nesting habitat. Our findings suggest that mixed-severity fire in ponderosa pine landscapes promotes native bee biodiversity.This work was supported by the McIntire-Stennis Cooperative Forestry Capacity Funding Program from the U.S. Department of Agriculture’s National Institute of Food and Agriculture (project award no. 7006470), the David R.M. Scott Endowed Professorship in Forest Resources, and the 2024 Garden Club of America Board of Associates Centennial Pollinator Fellowship
The Convex Algebraic Geometry of Higher-Rank numerical Ranges
No full textThesis (Ph.D.)--University of Washington, 2026The higher-rank numerical range is a convex compact set generalizing the classical numerical range of a square complex matrix, first appearing in the study of quantum error correction. In this thesis, we will discuss some of the real algebraic and convex geometry of these sets, including a generalization of Kippenhahn’s theorem, and describe an algorithm to explicitly calculate the higher-rank numerical range of a given matrix. We will also discuss the inverse field of values problem, an inverse problem on the numerical range. We focus on the geometric properties of the set of solutions. Finally, we consider an analogous problem for higher-rank numerical ranges and show how to solve it using the ideas behind the proof of convexity for these sets
Benchmarking and Advancing Knowledge Gap Navigation
Full text linkMost language-based assistants follow a reactive ask-and-respond paradigm, requiring users to explicitly state their needs. As a result, relevant but unexpressed needs often go unmet. Existing proactive agents attempt to address this gap either by eliciting further clarification, preserving this burden, or by extrapolating future needs from context, often leading to unnecessary or mistimed interventions. We introduce PROPER, Proactivity-driven Personalized agents, a novel two-agent architecture consisting of a Dimension Generating Agent (DGA) and a Response Generating Agent (RGA). DGA, a fine-tuned LLM agent, leverages explicit user data to generate multiple implicit dimensions (latent aspects relevant to the user’s task but not considered by the user) or knowledge gaps. These dimensions are selectively filtered using a reranker based on quality, diversity, and task-relevance. RGA then balances explicit and implicit dimensions to tailor personalized responses with timely and proactive interventions. We evaluate PROPER across multiple domains using a structured, gap-aware rubric that measures coverage, initiative appropriateness, and intent alignment. Our results show that PROPER improves on quality scores and win rates across all domains, achieving up to 84% gains in single-turn evaluation and consistent dominance in multi-turn interactions
Presence Scale Technical Report: Iterative Conceptualization, Psychometrics, and Validity Evidence
Full text linkThis technical report details the four-year process through which we developed and completed an initial validation of the Presence Scale (see Appendix K or Table 20 for the finalized scale). The Presence Scale is a 14-item self-report measure of Presence, a hierarchical construct composed of three factors: Stillness of Mind, Present Moment Awareness, and Consciousness Beyond Self. Our goal in creating the Presence Scale was to have a self-report measure from which we can draw inferences about the degree to which a person experienced Presence during a recalled experience or, eventually, following an experimental condition or immediately after an experience in everyday life. Our formal definition of Presence and its three factors are provided below:
Presence is the state of being in which conditioned thinking ceases, and the mind is open, aware, non-reactive, and still; and often a witness of itself. In this state, one can experience one’s consciousness as expanding, and one’s self as becoming part of something larger than the self. Presence is the mind attuned. Subject, not object. Life affirming.
Stillness of Mind occurs when discursive thinking subsides and the mind is clear, calm, and settled.
Present Moment Awareness occurs when one is open to and aware of the now, of the present moment, of being, even if one is involved in activity.
Consciousness Beyond Self occurs when one’s consciousness seems to expand beyond the confines of the body and mind, and potentially the self merges—one experiences becoming One—with another entity or realm.
The formal presentation of Presence and the core empirical evidence supporting the validity of the Presence Scale are reported in Kahn et al. (in press). This technical report provides the comprehensive methodological and empirical foundation underlying those findings as well as supplemental validity evidence. Here, we document the entire scale development process, including conceptual iterations, pilot studies, factor analyses, and item-level psychometric evaluation that extend beyond the scope of the journal article (see Figure 1 for an overview). Through providing transparent information about our scale development process, this report serves as a resource for researchers who wish to evaluate the Presence Scale and implement it with fidelity in their own work
How Washington’s Transition to Kindergarten program addresses childcare deserts: Current state funding caps and proposed funding cuts exacerbate gaps in access to early learning and disproportionally harm higher-poverty communities
Full text linkEarly childhood care and education (ECE) plays a critical role in supporting childhood development, providing children with fundamental rights that are outlined in the United Nations Convention of the Rights of the Child. Moreover, research shows investments in ECE provide long-term benefits for the individual and for society that far outweigh societal costs. Effective ECE systems provide caregivers with the opportunity to join the workforce if they choose, providing additional economic benefits to society. Washington state operates two state-funded ECE programs, the Early Childhood Education and Assistance Program and Transition to Kindergarten, or TK. Governor Ferguson’s proposed 2026 budget cuts funding for TK, a move that would have devastating effects on the state’s ECE system, especially within the most under-resourced areas of the state. This brief explains the important role that TK fills in addressing childcare deserts within Washington’s ECE system. We demonstrate that districts located in childcare deserts rely more heavily on TK. The brief offers policy recommendations for local and state policymakers
Scalable Data Paradigms for Steering General-Purpose Language Models
No full textThesis (Ph.D.)--University of Washington, 2026Pretrained Language Models (LMs) have demonstrated remarkable general-purpose capabilities by encoding vast amounts of knowledge from the internet. However, effectively steering these models to serve diverse downstream applications, such as following instructions, chatting with users, using tools, or performing complex reasoning, poses another set of challenges that require diverse, high-quality, and increasingly costly training data. This dissertation explores scalable paradigms for structuring, creating, and optimizing data to facilitate the broader generalization of language models and enhance their critical capabilities.First, through the creation of the SuperNaturalInstructions benchmark—a large-scale dataset with over 1,600 NLP tasks—I demonstrate that unifying NLP tasks via natural language instructions enables model generalization at the task level. Second, I propose Self-Instruct, a novel framework where LMs generate their own instructional data to train themselves, thereby demonstrating model self-improvement. Third, I develop HyPER, a framework that routes preference annotation tasks between humans and AI to optimize data quality and collection efficiency for preference-based learning. Finally, I systematically study the impact of diverse open instruction-tuning datasets on LM capabilities, leading to the development of the Tülu series of openly available and highly capable models. Together, these efforts—unifying task structures, leveraging model-generated synthetic data, optimizing human-AI data partnerships, and fostering open data ecosystems—have demonstrated an effective path to building a strong, scalable, and community-driven data foundation for post-training language models. Finally, I envision future directions that can further enhance this data foundation for building more advanced and sustainable AI systems
Brain Poetics: Temporal Perception and Reading 19th Century Poetry
No full textThesis (Ph.D.)--University of Washington, 2026Combining current cognitive science on time perception with Victorian poetry, this dissertation investigates how poetry alters the subjective time of a reader. Since poetry is a durational art form, with poems unfolding in time as the reader moves through the words within the poem, the time encoded in the language of the poem is transmitted to the reader as modulation in their own temporal perceptions. This power is enacted primarily through the formal structures of poetry that underlie the content of the poem itself: meter, rhyme, rhythm, alliteration, anaphora, and other methods of turning words into temporal patterns. These formalist considerations reinscribe understandings of what poetry is expected to do to a reader by carrying meaning, emotion, and, now, time. Three models are developed for the implications of time perception: (1) entrainment and how the real time processing of language allows formalisms to be felt in the reader's body, (2) the psychological present and how a more robust understanding of "now" enables explanations about the poetic language feeling alive, and (3) neuro-genre as an application of the brain's own genre schema for recognizing poems and how this complicates debates about the nature of the lyric genre. This final intervention tests definitions of the lyric poetry genre by placing it under the real-world conditions of nineteenth-century newspaper reprinting to demonstrate how print culture gives additional evidence for these methods of understanding reading. While Victorian Studies has considered how the history of science impacted literature, this dissertation allows for the current science of temporal perception to add to understandings of why poetry captures readerly attention and generates meaning for so many readers
Essays on Consumer Preferences and Demand for Aquaculture and Wild Fish
No full textThesis (Ph.D.)--University of Washington, 2026This dissertation comprises three empirical chapters investigating U.S. consumers' perceptions and demand for aquaculture products within the broader animal protein market. Using both revealed and stated preference methods, I analyze how environmental concerns, production methods, and ecological shocks influence consumer behavior, while developing improved demand forecasting tools for perishable protein products. \textbf{Chapter 1} addresses a fundamental gap in the animal protein literature by examining consumer choices across fish, chicken, and beef simultaneously, reflecting the real-world choices set for consumer. Based on survey data from 1,274 consumers across seven Seattle farmers markets (June–September 2019), we estimate willingness-to-pay (WTP) for key product attributes using a mixed logit model. The results reveal a clear asymmetry in production method preferences: free-range terrestrial proteins command a premium of \\beta = 0.019p<0.001\beta = -0.014p\approx0R^2R^2$ of 0.23. In percentage terms, this corresponds to a reduction of forecast error by nearly 70\% relative to baseline seasonal naïve models. An anomalous finding is that a simple median forecast produced the lowest MAPE (2.24\%), outperforming even tuned XGBoost (2.96\%), reflecting the bias of percentage-based metrics in intermittent demand settings. Feature importance analysis confirms the economic relevance of predictors: price and promotions drive short-term fluctuations, lagged sales capture stockpiling and purchase regularity, brand and store identifiers reflect consumer heterogeneity, and seasonal indicators align with calendar-driven consumption cycles. These findings advance the forecasting literature by linking machine learning outcomes with economic interpretation, while also providing actionable insights for retail managers and policy makers. Improved forecasts can enhance pricing and promotional strategies, reduce food waste through better inventory management, and support more sustainable supply chains. The study concludes by noting key limitations—including missing causal variables, evaluation metric biases, and context-specific scope—and outlines future directions involving richer data integration, SHAP-based interpretability, and pipeline approaches to study cross-product substitution effects. Collectively, this research advances understanding of aquaculture demand dynamics, quantifies the market impacts of ecological narratives, and delivers practical forecasting improvements for perishable animal protein products supply chains