71446 research outputs found
Sort by
A Balancing Act: Examining the Intersection of Data Privacy Rights and the Detection of Child Sexual Abuse Materials
No full textNAComputer Scienc
Question-Driven Reasoning in AI-Assisted Decision-Making: A Content-Based Approach
No full textWe introduce the Question-Driven Theory of AI-Assisted Decision-Making, a content-based model of reasoning to explain human-algorithm interaction in high-stakes risk assessment contexts. According to the content-based theory of reasoning proposed by Koralus et al., human reason is driven by the goal of reducing the complexity of our questions as directly as possible, as opposed to alternative reasoning models that focus on maximizing expected utility by considering all possible alternatives. By emphasizing the role of the question-answering mechanism in human reasoning, this theory allows us to bring a new perspective to theoretically model and thereby understand the dynamics of human-algorithm interaction. According to our theory, AI-assisted decision-making can be understood in two phrases, the Question-Raising phrase and the Question-Answering phase. First, the algorithm prediction guides the decision-maker to raise an actionable question, while adding dependencies of concepts into their mental model representation. Then, the decision-maker reaches a decision by settling the question with their mental model representation. We propose the design of and conduct a behavioral experiment to test the applicability of a content-based theory on AI-assisted decision-making. We show that the presentation of an algorithm’s risk assessment predictions, such as logically equivalent information with either positive or negative probabilities, can influence the type of questions decision-makers pose and, subsequently, the decisions they make. We propose a framework to computationally model this procedure, as well as normative directions for designing AI assistance that prompt the right kind questions to ensure rational decision-making procedure and fair results.Computer Scienc
Redesigning Opportunity: Creating Visual Tools for Low-Income Parents to Understand Socioeconomic Mobility
No full textThis thesis addresses declining socioeconomic mobility in America by developing a user-centered
visualization tool for low-income families. Despite research showing how geography impacts children’s outcomes, existing tools like the Opportunity Atlas primarily serve researchers rather than
families making critical decisions. Through a six-phase approach—need-finding interviews, thematic analysis, prototype development, co-design, enhancement, and testing—we created OpportunityViz, a web application translating mobility data into accessible guidance. Through interviews
with first-generation, low-income parent-child pairs, we identified ten themes shaping mobility
experiences and developed corresponding features. Usability testing showed features providing personalized recommendations, cultural considerations, and concrete action steps were highly effective.
Findings reveal effective mobility tools must offer personalization, cultural sensitivity, actionable
guidance, multiple pathways, and language accessibility. This research demonstrates how technology can bridge academic knowledge and practical application for vulnerable populations. By making mobility information accessible, OpportunityViz can empower disadvantaged families to make
informed decisions about their children’s futures, contributing to greater educational equity. A
demo of the tool is available at https://opportunity-ai-omega.vercel.app/Computer Scienc
Attending and Fellow Perspectives on Pediatric Hospital Medicine Fellow Supervision
No full textBackground:
Pediatric Hospital Medicine (PHM) is one of the newest pediatric subspecialties, and the number of PHM fellowship programs continues to grow rapidly. Many PHM attendings have limited experience supervising fellows. Currently, the PHM fellowship program at Boston Children’s Hospital lacks a structured curriculum to prepare attendings for clinical supervision. This study aims to conduct an organizational needs assessment among PHM attendings and fellows to identify faculty development needs related to fellow supervision. This assessment represents the second step in Kern’s six-step approach to curriculum development, as incorporating faculty and fellow input is essential for designing a curriculum that meets attendings’ needs.
Methods:
This cross-sectional study utilized three surveys targeting: (1) attendings at the main academic site (BCH), (2) attendings at the community site (SSH), and (3) current and graduated PHM fellows. The attending surveys assessed overall preparedness for fellow supervision, key topics to include in a curriculum, and overall interest in a curriculum. Fellow surveys explored their experiences working with attendings. Surveys were distributed anonymously via Qualtrics. Data analysis included descriptive statistics and a modified qualitative analysis using the framework method for free-response questions.
Results:
Results from BCH and SSH surveys varied given the inherent differences in training environments. The academic site (BCH) had a survey completion rate of 71.1% (32/45). Attendings expressed a need for more training on supervising both residents and fellows while tailoring teaching to the fellow’s level. Additionally, the majority of attendings at both sites shared interest in participating in a curriculum on fellow supervision. At the community site, the completion rate was 53.8% (7/13), with attendings identifying teaching to the fellow’s level and providing feedback as key training needs.
The fellow survey had a completion rate of 78.6% (11/14). Fellows identified promoting fellow autonomy as a critical topic at both training sites. Qualitative responses highlighted the need to educate community site faculty on the overall goals of fellowship training. Across all groups, the preferred curriculum delivery method was a shared document or resource folder.
Conclusion:
Our organizational needs assessment highlights the need for enhanced attending training on PHM fellow supervision from the perspectives of both attendings and fellows. While training priorities may differ between the academic and community sites due to the inherent differences in training environment, ongoing faculty development is essential for effectively preparing the next generation of PHM fellows.Graduate Educatio
Deficiency of Sensory Afferents Impairs Dental Pulp Stem Cells (DPSCs) in a Mouse Model of Pulpitis
No full textVital pulp therapy has recently gained traction within the endodontic community as a more conservative and biologically driven alternative to traditional root canal treatment. While sensory nerves are primarily known for their role in pain perception, emerging evidence suggests they also contribute significantly to tissue repair by regulating dental pulp stem cells. However, the precise mechanisms underlying the interaction between sensory nerves and dental pulp stem cells during injury repair remain poorly understood. This study investigates whether sensory nerve deafferentation affects DPSC-mediated homeostasis and the injury repair process in pulpitis. Using a Gli1-CreER/Rosa26-tdTomato mouse model, we performed inferior alveolar nerve deafferentation followed by a dental pulp capping procedure. Single-cell RNA sequencing post-deafferentation revealed a phenotypic shift in Schwann cell clusters, transitioning from a non-myelinated to a myelinated state. Additionally, a significant increase in immune cell populations was observed, indicative of heightened pulpal inflammation following sensory nerve loss. Importantly, sensory nerve deafferentation led to a marked reduction in Gli1⁺ cell aggregation and impaired tertiary dentin formation at the site of pulp capping. These findings suggest that sensory nerves play a pivotal regulatory role in guiding the migration and differentiation of Gli1⁺ cells into odontoblast-like cells, which are essential for effective tertiary dentinogenesis. This study illuminates a previously underexplored dimension of vital pulp therapy, emphasizing the potential role of sensory afferents and their associated molecular mediators in orchestrating mesenchymal stem cell behavior—particularly their migration and transdifferentiation—during the reparative response to pulpal injury.Endodontic
Essays on Institutions, Beliefs, and Asset Prices
No full textThis dissertation comprises three chapters that examine the effect of institutional characteristics and preferences and investor beliefs on asset prices. I show that mixed bond mutual funds transmitted Quantitative Easing to the corporate bond market, dissect the composition of investor disagreement, and propose a risk premium resolution of the FOMC bond premium puzzle.
In the first chapter, I study the transmission of Large-Scale Asset Purchase (LSAP) through financial investors’ balance sheets, illustrating the channel through US corporate bonds. Post-2008 LSAPs didn’t directly target corporate bonds, but cross-market investors might transmit the shock to corporate bonds through their portfolio adjustment behaviors. I show that mixed bond mutual funds, who I label as “switchers”, accommodated all three rounds of QE by selling Treasury securities and agency MBS. During the same period, “switchers” increased their corporate bond holdings by a total of 254.7 billion dollars. For two of the three rounds of QE, this switching behavior is associated with detectable price impacts on the corporate bonds that they held: for otherwise similar corporate bond issues, one percent higher ownership by “switcher” funds predicts 2.6 (1.0) basis points lower yield, eight quarters after the onset of QE2 (QE3). The effect is stronger at the firm level and concentrated in firms with speculative-grade ratings: for otherwise similar speculative-grade firms, one percent higher firm-wide bond ownership by “switcher” funds predicts 8.0 (8.4) basis points lower yield, eight quarters after the onset of QE2 (QE3). I show that these effects on risky asset prices affect firm borrowing decisions, as firms with higher “switcher” ownership had higher liability ratio and probability of issuing bond after QE2 and QE3. These effects on bond prices and firm borrowing decisions are not present in QE1 and placebo periods.
In the second chapter, which is joint work with Robin Greenwood and Sam Hanson, we ask an empirical question: when investors disagree about the prospects of a firm, is it because they disagree about the industry or because of different assessments of the winners and losers in that industry? We decompose analyst disagreement about future EPS, Sales and Long-Term-Growth (LTG) into an idiosyncratic and an industry component. We show that the majority of disagreement is driven by the idiosyncratic component, meaning that investors mostly agree about industry prospects but disagree about which firms will perform the best within each industry. We present a model in which even idiosyncratic disagreement can have industry and market valuation effects. Even if investors agree on the prospects of the industry, in the presence of short-sales constraints the industry may be overvalued when investors disagree about the prospects of individual firms. We find evidence in future returns consistent with this idea.
In the third chapter, I study the FOMC announcement premium for long-duration bonds, which is the fact that the average price return on 10-year nominal US Treasury securities during an FOMC announcement window (day before, day of, and day after an FOMC announcement) is 39 to 68 times that of a normal trading day (Hillenbrand [2023]). I present a model where risk aversion from a segment of the market (nervous sellers) drives the FOMC bond premium. Consistent with model predictions, bond returns are lower (not statistically significant) 2 days to 10 days before FOMC announcements. In addition, consistent with the model, the FOMC bond premium is higher when VIX is higher, realized variance of past bond returns is higher, and analyst forecast dispersion is higher. Using market-making data in the cash bond market from a large primary dealer, I examine investor flows around FOMC announcements. I find weak evidence that a subset of investors are persistent sellers of bonds before FOMC announcements, but the magnitude of nervous selling is too small to account for the FOMC bond premium.Business Economic
In the Blink of an Eye: A Unified Theory for Feature Emergence in Generative Models
No full textGenerative models, which produce samples of data such as text or images, are transforming the way we interact with technology. However, they often fail quickly in problematic and unintuitive ways. For example, a language model given a software engineering problem suddenly switched from coding to searching for pictures of Yellowstone National Park, and these rapid shifts in behavior have been observed in reasoning traces and hacks. This phenomenon is not unique to language models: in image generation models, key features of the final output, like objects in the background or the color, are also decided in narrow “critical windows” of the generation process.
While critical windows for a particular type of image generation model called diffusion have been studied at length by statistical physicists, existing theory relies on the specifics of diffusion and strong assumptions on the distribution of model generations. In this thesis, we develop a unifying framework for critical windows that shows that they emerge generically when the sampler specializes to a sub-population of the distribution it models. Drawing on tools from information theory, machine learning, high-dimensional probability theory, and statistical physics, our theory improves upon previous work by using rigorous mathematical tools and is agnostic to the underlying model type or distribution, applying to both language models and diffusion. The key insight of our approach is to exploit the powerful formalism for generative models of stochastic localization, which has roots as a proof technique in probability theory.
Leveraging our consolidated theory for critical windows, we apply it to different examples of critical windows in theoretical and empirical contexts. We provide a novel interpretation of the all-or-nothing phase transition in statistical inference as a critical window and use our framework to explain different failure modes of language models. We finally validate our predictions empirically for real-world models, and demonstrate that critical windows have applications towards improving the safety, privacy, and fairness of generative models.Computer Scienc
Introspective Discrimination: Probing the accuracy of memory, metacognition, and psychobiological prediction models under negative emotional contexts
No full textEmotional experiences emerge from a complex amalgamation of objective and subjective information. However, due to the partially subjective nature of emotional experiences, objective and subjective measures of experience can sometimes dissociate. These dissociations can consequently affect the accuracy of memory, emotional well-being, and the conscious perception of emotional experiences. In the following body of work, I leverage a combination of statistical modeling, computational modeling, and machine learning alongside behavioral experiments and observational studies to provide insights into the alignment between objective and subjective measures of emotional experience. In study 1, I examined how objectively accurate autobiographical memory is for subjective emotional experiences. The results of this study conclusively demonstrate that autobiographical memories of subjective emotional experiences are objectively inaccurate in negatively valenced contexts, and that overestimating the subjective intensity of these remembered experiences is common and negatively associated with current emotional well-being. Study 2 examined how objectively accurate memory and associated subjective confidence judgements are under emotional contexts. Insights from this experiment suggest that negative valence influences how we think and respond – objectively biasing actions, and influences our self-monitoring capabilities – subjectively biasing self-confidence. Study 3 further develops this foundation of knowledge to examine the predictive validity of objective physiological information for predicting subjective reports about the intensity of emotional experiences. This study demonstrates that changes in electrodermal activity (a measure of physiological arousal) can effectively track the intensity of valenced experiences, but fail to capture nuanced variations of specific emotional states.
In other words, how we consciously feel and label our emotions is more complex than what can be measured by changes in physiological arousal alone. Altogether, this dissertation furthers our knowledge of the association between objective and subjective measures of emotional experience and begins to reveal how negative emotional contexts are linked to dissociations between these constructs – ultimately affecting emotional well-being, the accuracy of memory and metacognition, and the perception of emotional experiences.Psycholog
The Role of Vision in Single-Leg Balance
No full textIt is something that we all take for granted, but humans’ ability to balance is a complex task. Upright stance is not a stable equilibrium, and without any control from our muscles, even the smallest disturbances would cause us to fall over. In this study, I looked into how the integration of different sensory feedback mechanisms allows us to properly balance when performing tasks such as walking, jumping, or standing on one foot. Specifically, I wanted to examine the role of vision in balance, and strategies we deploy to balance without it. When we close our eyes, how does our body make up for this loss of information, if at all? In order to better understand this question, we had participants stand on one leg with their eyes open and closed, and their preferred balancing strategies for each trial were tracked. It was observed that when subjects had their eyes open, they preferred to use the ankle strategy, in which one actuates their ankle in order change their center of pressure. Without vision, the subjects incorporated the hip strategy in concert with the ankle strategy, oscillating their torso side to side in order to stabilize their center of mass. I then compared this experimental data to an inverted pendulum model in order to see if I could isolate the role of vision from other feedback, and saw that a strategy analogous to our subjects’ hip strategy was employed once the model was unable to determine its absolute height. These findings are a first step to help us understand exactly how the sensory information we receive is reflected in our behavior, and points us towards interesting questions in human sensory integration which can help better inform sensory processing disorders.Applied Mathematic
Variation and Therapeutic Potential of Immune Cell States
No full textThe immune system plays a pivotal role in safeguarding the body by targeting pathogens, parasites, and abnormal self-cells, including cancerous ones. Immune cells exhibit significant heterogeneity, which is shaped by their development during hematopoiesis, signaling environments, tissue contexts, and disease. This diversity offers numerous opportunities for therapeutic interventions. This dissertation explores the interplay between immune cell variability and therapeutic potential across diverse contexts, emphasizing their roles in hematopoiesis, cancer immunotherapy, and plasma cell dyscrasias.
In the context of cancer immunotherapy, we dissect the cellular and molecular mechanisms driving immune-related adverse events (irAEs) and therapeutic efficacy. Using anti-CD40 immunotherapy in mice, we demonstrate that TH1-promoting cytokines IL-12 and IFN-γ are central to liver toxicity, with Kupffer cells and neutrophils driving inflammatory damage, while dendritic cells and CD8+ T cells mediate tumor control. These findings illuminate pathways to decouple immunotherapy efficacy from associated toxicities, providing strategies to mitigate irAEs while preserving therapeutic benefits. Separately, an in-depth analysis of neutrophil heterogeneity reveals a Sellhi neutrophil subset with an ISG signature whose infiltration is linked to tumor control in immunotherapy. This study provides strategies to harness neutrophils to enhance immunotherapy outcomes.
We further investigate the bone marrow as a site in both health and disease, focusing on plasma cell dyscrasias such as AL amyloidosis. By combining single-cell RNA sequencing, clonal profiling, and microenvironmental analysis, we uncover unique transcriptional states and niche dynamics associated with disease. Complementing this, we present an early effort to construct an atlas of hematopoietic variation across healthy individuals, revealing demographic and environmental influences on immune cell composition. Together, these studies bridge fundamental immunology with clinical applications, advancing our understanding of immune cell variability in health, disease, and therapeutic innovation.Systems Biolog