Alliance One Tobacco (Malawi)
Academic Research Repository at the Institute of Developing EconomiesNot a member yet
52305 research outputs found
Sort by
The American Ritual of Violence: A Cultural Analysis of Mass Shootings at Schools in the United States
No full textThis thesis uses anthropology to explore why young white men commit mass
shootings in schools in the United States in an effort to prevent future mass shootings.
Existing scholarship on the topic comes from the fields of psychology, sociology, and
criminology. While anthropologists study violence more generally, there has been limited
consideration of violence in the context of school shootings. In an attempt to further the
anthropological exploration of why young white men commit mass shootings in schools,
this project concentrates on three areas: Foucauldian discourse analysis on hegemonic
masculinity content created by school shooters; the ritualization of the process to commit
a school shooting informed by Roy Wagner’s scholarship on ritual theory analysis; and
an exploration of the body as a site of meaning when paired with a gun in the creation of
hegemonic masculinity. Evidence considered includes the online posts of school shooters
from message boards, YouTube commentary sections, websites, and remnants of their
journals and manifestos. The discourse was coded into primary themes that revealed the
young men who became shooters experienced social isolation, felt attacked by society,
and spent an overwhelming amount of time imagining violent massacres. We must offer
these young men opportunities to reintegrate into society before their feelings of
victimization set in and they view violence as their only pathway to hegemonic
masculinity. It might require writing a new origin story of the American man, one that
does not solely depend on the exercise of violence and power to both identify and elevate
itself above all others at any cost.Extension Studie
Combining Analytical, Numerical, and AI Models to Improve Cloud and Convection Representation in Climate Simulations
No full textAccurate representation of small-scale processes, such as convection and cloud formation, remains one of the greatest challenges in climate modeling, even in kilometer-scale storm-resolving simulations. These processes are essential for determining large-scale atmospheric behavior, but computational constraints prevent their full representation in global climate models. For instance, deep convection may not exhibit convergent behavior with increasing resolution, and significant uncertainties persist in ice microphysics parameterization. This thesis explores these challenges in two parts: the first (Chapters 2 and 3) examines how small-scale convection and cloud processes influence large-scale atmospheric states in idealized radiative-convective equilibrium simulations, while the second (Chapter 4) investigates machine learning (ML) as a tool for efficiently emulating these processes in climate models.
Chapter 2 addresses a fundamental question: what controls the vertical thermal structure of an equilibrium atmosphere? To answer this, I developed a refined zero-buoyancy plume model that analytically solves equilibrium atmospheric profiles given boundary conditions. The model highlights how plume-environment mixing influences vertical temperature profiles, upper-tropospheric convective mass flux, and cloud fraction. These findings align with convection-permitting simulations, which reveal that higher horizontal resolution---acting as a proxy for enhanced plume-environment mixing---leads to increased cloud fraction and mass flux in the upper troposphere.
Chapter 3 explores the impact of microphysics scheme uncertainties on equilibrium atmospheric states, particularly focusing on deep convective overshoots into the tropical tropopause layer (TTL). We find that different microphysics schemes produce distinct heat balance regimes in the TTL. Two schemes lead to a “hard-landing” scenario, where frequent, strong convective overshoots induce significant cooling (~0.2 K day−1), while a third scheme results in a “soft-landing” scenario, with weaker overshoots and minimal cooling (~0.03 K day−1). This difference arises from variations in upper-tropospheric stratification driven by atmospheric cloud radiative effects (ACRE). The scheme producing the soft-landing scenario generates stronger ACRE, leading to a ~3K warmer, more stable upper-tropospheric layer that buffers convective updrafts.
Chapter 4 demonstrates how ML can emulate these small-scale processes efficiently by learning directly from high-resolution simulations. Using data from superparameterized climate simulations, we train ML models to replace the embedded cloud-resolving models. While previous studies show that such hybrid ML-physics simulations can reproduce key climate statistics, they often suffer from online instability, particularly in setups with real geography and explicit cloud condensate coupling. By integrating an expressive U-Net architecture with cloud microphysics constraints, we achieve stable and skillful multi-year hybrid climate simulations with realistic cloud climatology and explicit cloud condensate coupling.Earth and Planetary Science
Famous Views of Economics: An AI-Based Analysis of Japanese Woodblock Print Subject Matter and Economic Circumstances
No full textHow do socioeconomic conditions influence what artists depict in their artworks? More concretely, how do artists react to trade shocks and technological innovation? To answer these questions, I leverage novel artificial intelligence (AI) methods to construct and study a bespoke dataset of Japanese woodblock prints, artist biographies, and international trade records from 1830-1939. I first detail how this novel AI methodology can unlock previously unavailable insights into contemporary culture and economy. Then, through regression discontinuity analysis, I show that artworks reflect "hype cycles" in specific object depictions around technological innovations and Japan's opening of trade in 1854 which precede the widespread proliferation of these novel objects. I then point to a potential mechanism of the artist's decision process as artists' depictions of foreign objects are correlated with their educational background and their geographic locale. Then, I connect artistic subject matter with economic variables to show that lagged artistic depiction of objects can be predictive of future imports, evidence that may suggest artworks could be proxies for consumer interest and eventual demand. Finally, I conclude by noting potential extensions of my AI-based approach.Applied Mathematic
A genomic history of the North Pontic Region from the Neolithic to the Bronze Age
No full textn/aThe north Black Sea (Pontic) Region was the nexus of the farmers of Old Europe and the foragers and pastoralists of the Eurasian steppe1,2, and the source of waves of migrants that expanded deep into Europe3–5. We report genome-wide data from 78 prehistoric North Pontic individuals to understand the genetic makeup of the people involved in these migrations and discover the reasons for their success. First, we show that native North Pontic foragers had ancestry not only from Balkan and Eastern hunter-gatherers6but also from European farmers and, occasionally, Caucasus hunter-gatherers. More dramatic inflows ensued during the Eneolithic, when migrants from the Caucasus-Lower Volga area7moved westward, bypassing the local foragers to mix with Trypillian farmers advancing eastward. People of the Usatove archaeological group in the Northwest Pontic were formed ca. 4500 BCE with an equal measure of ancestry from the two expanding groups. A different Caucasus-Lower Volga group, moving westward in a distinct but temporally overlapping wave, avoided the farmers altogether, and blended with the foragers instead to form the people of the Serednii Stih archaeological complex7. A third wave of expansion occurred when Yamna descendants of the Serednii Stih forming ca. 4000 BCE expanded during the Early Bronze Age (3300 BCE). The temporal gap between Serednii Stih and the Yamna expansion is bridged by a genetically Yamna individual from Mykhailivka in Ukraine (3635-3383 BCE), a site of uninterrupted archaeological continuity across the Eneolithic-Bronze Age transition, and the likely epicenter of Yamna formation. Each of these three waves propagated distinctive ancestries while also incorporating outsiders during its advance, a flexible strategy forged in the North Pontic region that may explain its peoples' outsized success in spreading their genes and culture across Eurasia3–5,8–10.Human Evolutionary BiologyAccepted Manuscrip
Attack of the Clones: Mapping melanoma clonal architecture using deep learning and single-cell multi-omics analysis
No full textMelanoma is an aggressive skin cancer marked by extreme genomic instability and resistance to therapy. As the disease progresses, clonal evolution becomes increasingly complex, driven by diverse genetic and epigenetic alterations. Traditional lineage-tracing approaches—such as copy number variation (CNV)-based phylogenetics—struggle to resolve this complexity, particularly in the context of spatial heterogeneity and convergent evolution. This thesis presents a new computational framework for clonal inference in melanoma, using mitochondrial DNA (mtDNA) mutations as stable, high-resolution markers of lineage.
In order to address the limitations of existing methods, I apply and adapt ReDeeM, a mitochondrial variant calling and lineage tracing pipeline, to melanoma. I further introduce a modified version of ReDeeM that incorporates UV mutation signature filtering, positional bias correction, and graph-based connectivity filters to reduce false positives. In parallel, I develop a graph neural network (GNN) model that constructs a cell–cell graph based on shared mtDNA, CNV, and chromatin accessibility features and helps infer downstream phylogenetic relationships through learned latent embeddings. Compared to CNV-based clustering, mtDNA-based clustering yields higher silhouette scores (0.46 vs. 0.32) and greater subclonal resolution. ReDeeM consistently identifies a broader set of informative variants, including low-frequency mutations missed by mgatk and other conservative callers.
The pipeline was applied to a single metastatic melanoma biopsy of over 9,000 cells. Using read-depth filtering and variant-level quality control, I identified robust mtDNA-based clusters and constructed phylogenetic trees with high internal consistency. Despite the tissue’s mutational burden and technical artifacts common in solid tumors, the approach revealed structured subclonal relationships with lineage coherence across multiple modalities. Modified ReDeeM and GNN methods produced consistent tree topologies and performed at least comparably to baseline methods in Adjusted Rand Index and phylogenetic distance metrics. Taken together, this work demonstrates that mtDNA-based clonal inference—augmented by careful filtering and graph-based modeling—may be able to resolve evolutionary structure in melanoma where other single-cell methods fall short. These results provide a foundation for tracking clonal selection under therapy and, ultimately, for anticipating resistance before it emerges.Computer Scienc
Language processing in the brain across diverse languages and speakers
No full textA staggering 7,000 languages are spoken and signed around the world. Furthermore, the majority of the world’s population speak two or more languages. Nevertheless, language research –especially in neuroscience – has been limited to primarily monolingual speakers of a couple dozen languages, which makes our understanding of how the language system works incomplete and potentially biased. In my dissertation, I use a precision fMRI approach to paint a more complete picture of language processing in the brain by studying both diverse languages and diverse speakers.
Language processing is supported by a set of inter-connected left-lateralized frontal and temporal areas that are jointly referred to as the ‘language network’. First, I asked whether core findings about this network that have come from studies of on native speakers of English and a few other languages generalize to speakers of typologically diverse languages. I found that four key properties of the language network – its i) topography, ii) left-hemispheric lateralization, iii) selectivity for language relative to non-linguistic inputs and tasks, and iv) strong functional connectivity are robustly present across 45 languages spanning 11 language families (Malik-Moraleda, Ayyash et al., 2022, Nature Neuroscience). I further extended these results to constructed languages (conlangs), such as Esperanto and Klingon. Constructed languages differ from natural languages in that they were created much more recently, often by a single individual and for diverse (sometimes esoteric) purposes, and are not shaped by learning and processing pressures, at least not to the same degree. I tested neural responses during the processing of five conlangs in proficient speakers, including Esperanto, Klingon, Na'vi, High Valyrian and Dothraki. All constructed languages recruited the same brain areas as natural languages, suggesting that constructed and natural languages share critical features despite their differences, thus constraining the definition of what constitutes a ‘language’ (Malik-Moraleda et al., 2023, under review).
Second, I investigated the neural architecture of bilingual and monolingual speakers. Contra claims that some frontal brain regions in bilingual individuals support both language processing and domain-general executive functions (e.g., Garbin et al, 2010, Coderre et al., 2016), I found no overlap between the language and the executive control networks in either monolingual speakers (replicating past work) or bilingual speakers (Malik-Moraleda et al., in prep). However, I found that bilinguals engaged the executive control network to a greater extent than the monolingual speakers during a demanding spatial working memory task (Malik-Moraleda et al., 2021, Neurobiology of Language). Although this result is in line with the controversial claim that bilinguals exhibit superior executive functions than monolingual speakers (Bialystok, 2001), it remains difficult to link it to bilingual experience specifically, rather than to other features that may differ between these populations.
Last, I investigated the language network of polyglots – individuals who have some degree of proficiency in five or more languages. In a group of 34 polyglots, including 16 hyperpolyglots with some knowledge of 10+ languages, I found that all languages—including the participants’ native language, non-native languages of varying proficiency, and even unfamiliar languages consistently engaged all areas of the language network. Moreover, the languages that participants rated as higher-proficiency elicited a stronger response, which plausibly reflects greater engagement of linguistic computations. The exception was the native language, which elicited a lower response than a non-native language of similar proficiency. These results contribute to our understanding of how multiple languages co-exist within a single brain (Malik-Moraleda, Jouravlev et al., 2024, Cerebral Cortex).Medical Science
Simulations of Judicial Pretrial Decisions to Explore Risk Assessment Validation
No full textThe pretrial phase in criminal cases is critical in determining the trajectory of a defendant’s case, particularly during the first appearance hearing when a judge decides whether to release or detain the defendant. Traditionally, judges made these decisions at their own discretion, but in recent years, data-driven risk assessment tools have emerged to guide these judgments. This thesis builds upon the growing body of research exploring the effectiveness of such tools, specifically focusing on the Public Safety Assessment, a widely-used pretrial risk assessment tool. Drawing from data sets from randomized controlled trials conducted by the Access to Justice Lab at Harvard Law School, I developed simulation code to explore the effects of risk assessment tools on judicial decisions and pretrial outcomes. This work provides researchers with a unique approach that uses simulations to examine how unobserved factors and missing data can influence the assumptions underlying risk assessment models.Applied Mathematic
The Effect on Extubation of Early vs. Late Definitive Closure of the Patent Ductus Arteriosus in Premature Infants: A Target Trial Emu-lation Using Electronic Health Records and A Target Trial Emulation Comparing High-frequency Jet Ventilation Management Strategies for Respiratory Acidosis among Neonates with Respiratory Failure
No full textIt’s well established that the key advantages of a randomized trial include: marginal exchangeability between groups, clear specification of time zero, and synchronization of eligibility check and treatment assignment with time zero. These features become even more apparent when considering the challenges of drawing causal conclusions from observational data. A reliable approach to maintaining the key attributes of randomized trials in observational analyses is to design them in a way that explicitly emulate a hypothetical randomized trial aimed at answering the causal research question — the target trial.
Premature infants are often referred for the definitive procedural closure of the patent ductus arteriosus (PDA) with the failure of, or contraindication to, pharmacotherapy and the inability to wean respiratory support. However, once this need is identified, the importance of expedited closure is unclear. We first specified a hypothetical randomized trial (the “target trial”) that would estimate the effect on extubation of early (0–4 days from referral) vs. late (5–14 days from referral) definitive PDA closure. We then emulated this target trial via inverse probability (IP) weighting, using a single-institution registry of premature infants (born weeks or with a birth weight 1500 g) who underwent the definitive closure of PDA between January 2014 and October 2023. The objective of this study was to compare the effect of the timing of definitive closure (i.e., surgical ligation or de-vice occlusion) on early respiratory outcomes in premature infants without complex congenital cardiac disease.
High-frequency jet ventilation (HFJV) is often used in neonatal intensive care units (NICUs) to treat respiratory failure. In neonates initiating HFJV for hypercarbic respiratory failure, clinicians aim to adjust ventilator settings to achieve a gradual reduction in pCO₂. Rapid correction with a large shift in pCO₂ may lead to cerebrovascular spasm and increase the risk of neurovascular injury. However, specific management of HFJV peak inspiratory pressure (PIP) varies across institutions. Here, we use EHR data and aim to compare two dynamic PIP management strategies by emulating a progmatic target trial. We illustrate the use of inverse probability (IP) weighting to adjust for time-varying confounding.Graduate Educatio
Logistics in the Line of Fire: A Stochastic Programming Model for Contested Environments
No full textAs military doctrine evolves to address the challenges of large-scale combat operations (LSCO), logistics planning must adapt to ensure the continuous sustainment of dispersed and contested forces. This thesis presents a two-stage stochastic mixed-integer programming (MIP) model designed to optimize military resupply under uncertainty. The formulation incorporates key doctrinal priorities, such as predictive logistics, prepositioned supply, and distribution network resilience, by explicitly modeling disruptions to supply routes and storage nodes. Using scenario data created from the Russo-Ukrainian War, the model evaluates the impact of adversarial attacks, storage costs, and vehicle availability on overall logistical performance. The results provide quantitative insight into the tradeoffs between transportation and storage-based supply strategies, revealing nonlinear thresholds that influence the model's behavior. The proposed framework offers a rigorous, extensible tool for analyzing logistics strategies in adversarial environments and contributes to the broader effort to formalize military logistics planning using mathematical modeling and data-driven decision making.Applied Mathematic