71446 research outputs found
Sort by
Convergent processing of auditory and tactile vibration in the inferior colliculus
No full textVibrations are ubiquitous in nature, shaping behavior across the animal kingdom. For mammals, mechanical vibrations acting on the body are detected by mechanoreceptors of the skin and deep tissue and processed by the somatosensory system, while sound waves traveling through air are captured by the cochlea and encoded in the auditory system. Here, we report that mechanical vibrations detected by the body’s Pacinian corpuscle neurons, which are unique in their ability to entrain to high-frequency (40-1000 Hz) environmental vibrations, are prominently encoded by neurons in the lateral cortex of the inferior colliculus (LCIC) of the midbrain. Remarkably, most LCIC neurons receive convergent Pacinian and auditory input and respond more strongly to coincident tactile-auditory stimulation than to either modality alone. Moreover, the LCIC is required for behavioral responses to high frequency mechanical vibrations. Thus, environmental vibrations captured by Pacinian corpuscles of the body are encoded in the auditory midbrain to mediate behavior.Medical Science
The evolution of neo-sex chromosomes in Australian honeyeaters (Aves: Meliphagidae)
No full textSex chromosomes display remarkable diversity across the tree of life, including extraordinary variation in their number, size, content, sex determining pathways, and rates of turnover, often even among closely related lineages. In birds, the Z and W sex chromosomes were thought to be stable and immune from evolutionary turnover events because they arose over 140 million years ago and appear to be present in all extant avian lineages. However, discoveries of neo-sex chromosomes, which constitute a type of turnover involving autosomal fusions with ancestral sex chromosomes, have begun to accumulate in various avian lineages. My dissertation focuses on describing the structure and phylogenetic distribution of neo-sex chromosomes in honeyeaters (Aves: Meliphagidae), investigating potential consequences of their formation in the context of gene expression, and finally examining their potential role in climate adaptation. In Chapter 1, I characterize the putative fusion and structure of the neo-Z chromosome with genomic data by creating a high quality long-read genome and comparing it with other avian genomes. In Chapter 2, I integrate cytogenetic data and whole genome resequencing data to validate the fusion and structure of both neo-Z and neo-W chromosomes and resolve the phylogenetic distribution of neo-sex chromosomes in this clade. I also conduct phylogenetic tests to determine the timing and extent of recombination suppression on the neo-W, which is a hallmark of sex chromosome evolution and over time results in degeneration of the W. In Chapter 3, I investigate the evidence for dosage compensation on the neo-sex chromosomes, which is a regulatory mechanism that can evolve in response to degeneration of the neo-W to restore ancestral gene expression levels. I find evidence of incomplete dosage compensation in both the ancestral and added region of the neo-sex chromosomes. In Chapter 4, I conduct genotype-environment association analyses in a honeyeater with neo-sex chromosomes distributed across an aridity gradient in New South Wales. I find the ancestral Z has more significant associations with climate than the added Z and unexpectedly discover a number of large outlier regions on autosomes, including a polymorphic inversion, that are highly associated with climate and likely facilitating local adaptation through recombination modification. Taken together, my dissertation leverages genomic, transcriptomic, and cytogenetic data to shed light on a novel avian sex chromosome system and explore effects of its formation.Biology, Organismic and Evolutionar
(re) Interpreting Visions, (re) ConceptualizingTime: An Akamba Hermeneutic of Syokimau na wathani wake (her visions) and consequential Akamba location within the very long yùa ya mundù (Anthropocene)
No full textIn the age of the Anthropocene, ecological crises are sewing destruction in tandem with (or perhaps by causing) social unrest, economic inequity, and political instability. One of the narratives the Akamba rely upon in the Anthropocene is that of the famous muthani (seer) Syokimau. Stories passed down from Akamba elder to Akamba younger recall how Syokimau foresaw and warned of the arrival of British colonialism and its devastating and dismantling influence and impact on the Akamba. Through this story, the Akamba share a collective memory of a time gone by, in which contemporary adversities and future uncertainty are suspended to recall a moment in time before the commencement of a prior world ending calamity. However, in straying from the consensus interpretation amongst the Akamba, this dissertation argues that Syokimau’s wathani (visions) must not be understood to speak exclusively to the crisis of 19th and 20th century British colonialism and its aftermath. Employing Jacob Ọlúpọ̀nà’s methodological theory of indigenous hermeneutics, this project embraces the non-dualism and contradiction inherent within Akamba Oral Tradition to first analyze multiple retellings of Syokimau’s life and wathani (visions); investigate 7th to 19th century Akamba inhabitants of Ukambani and their participation in the Indian Ocean/Swahili trade as elephant hunters and slave traders; demonstrate, using various factors of Akamba culture including the Akamba language Kĩkamba, Akamba conceptions of time, Akamba clan system, Akamba storytelling, and Akamba rituals, that Akamba participation in a global trade rooted in extraction and exploitation was indicative of an Akamba community violating their syĩtheo (traditions); and finally engage with Sylvia Wynter’s theory on the “invention of Man” to contribute an Akamba commentary on the role of the Akamba in anthropogenic climate catastrophe.African and African American Studie
Computationally Speaking: The Mathematical Foundation of Large Language Models and An Exploration Into How They Tell Stories
No full textOver the past year, large language models such as ChatGPT have gained immense popularity, with hundreds of millions of active users. The adoption of these models in everyday tasks marks a significant shift in how we perceive and interact with tech- nology, making it all the more crucial to understand how these new tools work. This thesis aims to elucidate the inner workings of large language models, starting from first principles. We begin with an introduction to foundational machine learning concepts. Next we analyze the underlying architecture of neural networks, focusing on the evo- lution from basic feed-forward networks to Recurrent Feed-Forward networks, Long Short-Term Memory networks, and most importantly Transformer networks. In this analysis, we highlight key components such as the residual stream vector space and at- tention block. We then explore the optimization algorithms used to train autoregres- sive Transformer networks, including deterministic gradient descent, stochastic gradi- ent descent, and Adam, with an emphasis on their convergence properties. Finally, we present current research on Transformer network interpretability, including an ongoing research project about differentiating storytelling modes in the popular large language model Llama2. This thesis underscores that the first step to using machine learning responsibly is to understand it mathematically.Computer Scienc
How has the Depiction of the Folkloric Figure of the Fairy Evolved?
No full textThis thesis examines the evolution of the folkloric figure of the fairy, tracing how various societies imagined, depicted, and represented fairies across time and space. Unlike previous academic works, this thesis argues that five key “evolution causations” contributed to collective understandings of the fairy’s characteristics, attributes, and capabilities. Paying special attention to early fairy depictions in Ireland, Scotland, and England; Victorian era depictions in Britain and America; and contemporary Pop Culture depictions in America, I will analyze this evolution in connection with how the folkloric figure has been presented in various forms during its evolution in the locations of Ireland, Scotland, Britain, and America. Some of these cited evolution causations have been major factors in the evolution of the depiction of the fairy figure, while others have been minor but have still had some effects on the evolution of the depiction of the folkloric figure. By examining these causations, several paths of evolution are identified as well as several depictions of the folkloric figure of the fairy. This approach is important because by looking at the folkloric figure of the fairy in this new way, this approach allows us to understand how the figure has been used by several cultures for identity purposes, as a tool to record cultural history, and for its ability to create cultural cohesiveness. We also are able to see how the figure was used in cross-cultural exchanges and how the figure was used by different cultures for their own purposes.Extension Studie
Causal Inference Beyond Standard Assumptions: Learning Policies and Treatment Effects in Complex Environments
No full textCausal inference aims to uncover cause-and-effect relationships from data and has seen widespread application and rapid methodological development across scientific disciplines. While recent methodological advances --- driven by increasingly rich and diverse data ---- has expanded the scope of causal analysis, practical applications often involve complexities that violate the core assumptions underlying standard approaches. These include lack of overlap between treatment and control groups, interference among units, and distributional shifts across populations. Addressing these challenges is crucial for ensuring the validity and reliability of causal conclusions in real-world settings.
This dissertation develops novel methodological frameworks for robust, efficient, and interpretable causal inference in complex environments. Each chapter addresses a unique violation of standard assumptions, with an overall focus on two key areas: policy learning and heterogeneous treatment effect estimation.
Chapter 1 considers safe policy learning in regression discontinuity designs, where treatment assignment is deterministic and requires robust extrapolation beyond observed data. Chapter 2 focuses on the evaluation and learning of individualized treatment rules under clustered network interference, where spillover effects exist and may vary across units within a cluster. Chapter 3 investigates the generalization of heterogeneous treatment effects with multisite data, where distributional shifts across populations challenge the validity of pooled or site-specific estimators.Statistic
Engaging the Next Generation: Connecting Students with Collective Impact Organizations
No full textOne common challenge that organizations engaged in collective impact (Kania & Kramer, 2011) face is a growing need to reach broader audiences and inspire new leaders to advance this work. This capstone explores strategies to expand student engagement with organizations practicing collective impact, through the lens of The EdRedesign Lab (EdRedesign) at the Harvard Graduate School of Education. I leveraged the organization’s senior fellowship program, which supports best-in-class leaders, to introduce students to EdRedesign and the collective impact field.
A pivotal moment came when survey data revealed that this emergent field’s vocabulary may create a barrier to entry for students and other professionals. Findings showed widespread unfamiliarity with key terms such as collective impact, cross-sector work, backbone organization, among others. To address this communication barrier, I launched an informal awareness campaign using storytelling to explain collective impact work, demystifying the vocabulary and increasing accessibility.
To generate student engagement, I leveraged my connections at Harvard through my roles as a doctoral student, first-year experience residential proctor, and fellow. These efforts contributed to the creation and launch of a new pilot program, the Cradle-to-Career Summer Fellowship, aimed at providing undergraduate students with summer experiential learning at a collective impact organization. The fellowship may serve as a model for organizations across the country to adopt in their efforts to engage undergraduate students.
I make concrete recommendations to address both technical and adaptive challenges (Heifetz, 1994) that organizations may encounter when implementing similar practices. I also identify implications for my leadership and the sector at large.
Through this engagement project, I experienced managing from the middle, aligning my workstyle with the organization’s culture, and deepening my understanding of the collective impact ecosystem. My experience as a resident allowed me to learn from a strategic-minded team, reflect on my leadership style, and practice my immunity-to-change goals (Kegan & Lahey, 2009). On a broader level, this work reinforced the importance of intentionally developing a culture of mentorship to scale leadership and ultimately increase impact.Educatio
Modern Warfare: The Impact of Social Media on Citizen Journalism in Interstate Conflicts involving Ukraine
No full textThis thesis explores Ukrainian citizen journalism on social media during the first three months of the 2022 Russian invasion. It aims to fill a gap in the research regarding the use of social media by citizens during domestic conflicts and interstate conflicts.
This study defines ‘citizen journalists’ in the social-media age by building on the existing literature. I identify them primarily as residents of a country, who use their personal social media accounts to publish original content that has news value. I identified types and strategies of citizen journalism by analyzing 50 X (formerly Twitter) accounts covering the period from February 24, 2022, to May 31, 2022. Additionally, I conducted personal interviews with citizen journalists, mediators, and media members, aiming to gather further insights to add to the quantitative findings.
This study found that Ukrainian citizens produce citizen journalism using social media to appeal to Western, English-speaking audiences. The journalists do this out of a sense of civic duty, combined with a desire to participate in some manner, while also aiming to combat online propaganda about the war. Editorial standards—both among those formally trained in journalism and those who are first-time content producers—are apparent in the major efforts taken to identify sources. Techniques such as photos and videos are expanded, especially following their use as fact-checking tools during Euromaidan. Such techniques are favored by international audiences that are the direct target of citizen journalists.
Audience responses confirm the hypothesis that the news landscape on social media is shifting in favor of personal content, including diary-style posts, indicating that the use of social media by citizen journalists empowers more frequent, personal contributions to news stories, which in turn is likely to impact the way Western, English- speaking audiences are informed.Extension Studie
Advancing Evidence-Based Maternity Care: Empirical Studies of Technologies, Policies, and Clinical Practices in the U.S. and Abroad
No full textWhile maternal and newborn health (MNH) outcomes have improved globally over the past couple decades, progress has been slow or stagnant in many low- and lower-middle-income settings. In Kenya, the maternal mortality rate remains well over five times the United Nations’ Sustainable Development Goal and high rates of severe maternal and neonatal morbidity persist. Meanwhile, the United States trails almost all high-income nations with respect to MNH care, including variable family planning support across states and minimal recent declines, if not increases, in rates of adverse outcomes like maternal and perinatal mortality. Amidst this backdrop, global maternity care utilization has been on the rise, with growing use of services ranging from patient support tools to contraception to advanced imaging. Accordingly, this dissertation expands the evidence base on 1) a patient-facing digital health tool implemented in health facilities across Kenya; 2) recent Medicaid payment policies regarding provision of immediate postpartum long-acting reversible contraception (LARC); and 3) the utility of electronic-health-record- (EHR) and machine-learning- (ML)-based risk stratification models to identify and guide management for individuals at high risk of adverse outcomes like late stillbirth.
In Chapter 1, jointly conducted with Wei Chang, Sharon Akinyi, Sarah Little, Catherine Gakii, John Mungai, Cynthia Kahumbura, Anneka Wickramanayake, Sathyanath Rajasekharan, Jessica Cohen, and Margaret McConnell, I describe a parallel arm cluster randomized controlled trial carried out in 40 health facilities in Kenya to evaluate the impact of a low-cost, digital health platform called PROMPTS. Developed by Jacaranda Health, a leading MNH nonprofit, PROMPTS consists of informational messages, appointment reminders, and a two-way clinical helpdesk. Using longitudinal surveys of participants, we find that individuals recruited from facilities offering PROMPTS exhibited modest but consistent improvements across the pregnancy-postpartum care continuum, including in knowledge, preparedness, routine and danger sign care seeking, newborn care, and postpartum care content. We identify notable advances in the postpartum setting, for both mothers and newborns, which has important implications for efforts to improve postpartum care quality in Kenya.
In Chapter 2, jointly conducted with Maria Steenland, Benjamin Sommers, and Jessica Cohen, I evaluate Medicaid payment policies in Georgia and New York aimed at expanding contraceptive choice through separate reimbursement of immediate postpartum LARC. Using statewide hospital discharge data from the Healthcare Cost and Utilization Project and an interrupted time series design, we find significant, post-policy increases in the provision of immediate postpartum LARC, with accompanying reductions in the rate of subsequent, short-interval birth. We also find coincident decreases in rates of immediate postpartum sterilization, suggesting some amount of substitution between highly effective contraceptive methods. In subgroup analyses of adolescent individuals, Hispanic individuals, and non-Hispanic Black individuals – all of whom have higher rates of unintended pregnancy – we identify larger than average increases in immediate postpartum LARC provision and more pronounced reductions in the rate of subsequent birth within 21 months. Ultimately, our findings suggest that the payment policies led to new use of LARC and likely some reduction in the rate of unintended pregnancy.
In Chapter 3, jointly conducted with Jessica Cohen and Mark Clapp, I describe the development of an ML-based risk stratification model to prospectively identify individuals at high risk of adverse perinatal outcomes. Using features readily available in the EHR of a large, Massachusetts-based health care system, we first demonstrate the improvement in risk stratification from supplementing traditional, clinical risk factors with additional clinical, geographic, and sociodemographic features. We then leverage model predictions to generate insights about utilization of outpatient antenatal fetal surveillance (AFS) and its association with rates of late stillbirth, across risk groups. We find that rates of AFS use increased with predicted risk, suggesting that providers conducted surveillance at least in part based on similar factors to our model. Finally, we prospectively identified a predictably high-risk set of pregnancies for which AFS was associated with a nearly 80% lower rate of late stillbirth, though – given our observational approach – we could not causally attribute the lower outcome rate to AFS. Descriptively, we find that high-predicted-risk individuals who did not receive AFS lived further from the hospital and got less prenatal care but had lower rates of nearly all traditional clinical risk factors, including at the end of pregnancy.Health Polic
Transient Pattern Formation in Biological Systems
No full textThis dissertation investigates transient pattern formation in biological systems through a combination of theoretical modeling, stochastic analysis, and simulation. Biological processes are often driven by local interactions and feedback mechanisms that give rise to complex, time-dependent patterns. Given the inherent heterogeneity and noise in living systems, traditional deterministic models are often insufficient to capture the full spectrum of behaviors observed in nature. Here, I develop and analyze different models that are robust to microscopic details while capturing essential dynamical features.
One focus of this dissertation is the study of reaction-diffusion phenomena in immune cell signaling. I investigate how neutrophils generate self-regulating, transient chemical waves that coordinate a rapid yet contained response to injury or infection. The models show that the interplay between activators and locally produced inhibitors can naturally limit the spatial extent of these signaling waves, providing a mechanistic basis for preventing overreaction in immune responses.
Further, I examine the role of mechanical stress in flow-driven pattern formation within porous media. By representing these media as dynamic networks in which individual conduits adapt through erosion and deposition, I identify critical thresholds that lead to distinct phase behaviors, such as channelization and homogenization. This work not only elucidates the feedback between fluid flow and structural evolution but also offers a simple approach to analyze complex networks.
By applying a similar strategy to biological networks, I explore the emergence of optimized biological flow networks. By integrating local mechanical sensing into growth dynamics, I derive conditions under which vascular systems naturally converge toward configurations predicted by Murray’s law—a hallmark of energy-efficient design observed in blood vessels, leaf venation, and even in the foraging networks of slime molds.
Finally, I extend the classical mutation models, exemplified by the Luria–Delbruck experiment, to regimes where the effective mutation rate is significantly higher through modern gene-editing techniques. By formulating discrete stochastic models, I reveal novel phase transitions in DNA break-and-repair dynamics and demonstrate how randomness in molecular events influences cell fate and population heterogeneity.Engineering and Applied Sciences - Applied Physic