71446 research outputs found
Sort by
An Unplugged Curriculum: Empowering All Middle Schoolers with Computational Thinking
No full textComputational thinking is a newly popularized concept, and many have argued for its place in the K-12 curricula. The pedagogical approach is one that creates discussion and debate on how to effectively educate the current and future generations. This thesis presents a novel way of teaching computational thinking skills, specifically to middle school students. The approach links computational thinking concepts to everyday life occurrences and utilizes a teaching strategy that combines aspects of Montessori, peer learning, and unplugged instruction. This method enhances students’ comprehension of computational thinking concepts, cultivates their long-term interest, and encourages their engagement with the material at a deeper level. It helps bridge the unfamiliarity gap many students have with computer science and computational thinking and introduces a new way of thinking about problems in this heavily computational world. A curriculum framework is provided along with sample units that can be used, edited, or iterated upon. With hopes to continuously equip students for the ubiquitous technological world around them, this work serves as a first step in a new direction.Computer Scienc
Implementation and Evaluation of an mOral Health Training Program for Community Health Workers in Kenya
No full textGiven the acute shortage of oral health professionals, the WHO Global Strategy for Oral Health adopted at the 75th World Health Assembly recommended the development and implementation of workforce models that enable sufficient numbers of adequately trained, motivated, and well-distributed health workers to provide oral health services through a collaborative and interprofessional network. The Consortium of Universities for Global Health (CUGH) provided guidance on competency-based oral health education and in 2023, the Harvard School of Dental Medicine in partnership with the WHO-AFRO developed the first competency-based m-oral health curriculum designed specifically for CHWs in Africa. This study evaluates the effectiveness of the pilot implementation of the mOral health curriculum in preparing CHWs to deliver oral health care to underserved communities and identifies barriers and facilitators to implementing the training program to inform future scalability.
A mixed-methods evaluation was conducted using a convergent parallel design. Quantitative data were obtained from the OpenWHO platform, and a structured course feedback survey was distributed to all registered course participants in English, French and Portuguese through the platform. Qualitative data were collected through focus group discussions with CHWs and designated trainers (DTs) selected through purposeful sampling in four pilot counties in Kenya. Observational data collected through participation in planning meetings and digital ethnographic analysis of stakeholder discussions, were used to triangulate findings and assess implementation dynamics. Descriptive statistics were applied to quantitative survey data, while qualitative data were thematically analyzed to identify key facilitators and barriers to implementation.
The online course enrolled 5,957 learners globally, with 2,403 earning a record of achievement. The feedback survey in English received 371 responses, French had 123 responses and Portuguese had 46 giving us response rates of 6.2%, 10.9% and 8.9% respectively. Survey responses demonstrated high engagement, with 90% of participants affirming the course’s relevance to their work. A total of 1,265 CHWs and 39 DTs participated in the pilot implementation in Kenya. Focus group discussions were conducted with 40 CHWs (10 from each county) and 18 DTs. Qualitative findings highlighted the program’s success in addressing a critical knowledge gap, increasing CHW confidence in oral health promotion, and fostering interprofessional collaboration and quantitative findings confirmed their high satisfaction with the course curriculum and features. Implementation challenges included digital access barriers, insufficient logistical and financial support, and highlighted gaps in integration of CHWs into formal referral systems.
The study provides evidence supporting the feasibility of integrating CHWs into oral health service delivery through digital training. Key facilitators included the flexibility of the online format, the effectiveness of the mOral health curriculum, and the strong peer support networks among CHWs and DTs. Challenges such as technological barriers, and a lack of formalized referral pathways must be addressed for effective scale-up. While Kenya had a strong strategic plan that guided this work with highly engaged stakeholders, strengthening implementation through strategic policy support, provision of offline training resources, and sustainable funding mechanisms will be critical for expanding this model across other low and middle income settings. These study findings contribute to the broader discourse on workforce capacity-building and the role of digital learning in advancing universal health coverage.Dental Public Healt
Three Essays on Improving Measurement in the Study of Early Childhood Policy
No full textAs early childhood has grown as a subject of academic study and policy intervention internationally, there is increasing demand for measures of early childhood development and the quality of early childhood care and education settings that function well across diverse settings. This dissertation presents three essays on improving measurement of these important constructs and defines improvement as better understanding the psychometric properties of assessments, designing strategies that enhance reliability and validity, and addressing logistical and practical constraints alongside psychometric ones.
The first essay introduces a human-centered framework for evaluating a longer assessment of early childhood development (ECD) and creating a shorter version that maintains good measurement properties. This framework considers statistical, conceptual, and practical facets of items to maximize reliability, reduce complexity, and the conceptual breadth of the construct being studied. The framework is then applied to the International Development and Early Learning Assessment (IDELA), an assessment of ECD for three- to six-year-old children used in over 90 countries worldwide, to propose a balanced short form.
The second essay compares direct and caregiver-reported measures of ECD to evaluate the convention that direct assessment is a superior form of measurement. Using two datasets from the creation of the caregiver-reported ECDI2030 tool used to assess ECD with millions of two- to four-year-old children worldwide, this study evaluated the performance of item-pairs and compares the functioning of a directly assessed item against a caregiver-reported counterpart. The results suggest that direct assessment may deserve its status as a preferred form of ECD assessment, but that many applications of ECD measurement may be appropriate with caregiver reported tools.
The third essay evaluates the psychometric properties of the International Development and Early Learning Assessment-Classroom Environment (IDELA-CE), an assessment of process and structural quality of ECCE settings. It examined the factor structure of the IDELA-CE and how well it aligns with the hypothesized structure of the tool across countries. It also examines sources of measurement error stemming from the rater observing the class and the particular lesson observed. The analysis found moderate support for the factor structure hypothesized by the tool's creators and that while most variance in observed scores is attributable to stable differences between classrooms, that day-to-day fluctuations should not be ignored.
The overarching theme of this dissertation is that measurement in early childhood is a study of tradeoffs. The three chapters highlight tensions in ensuring that the validity & relevance, reliability & precision, and feasibility & cost of measurement is tailored to the use of scores.Educatio
Grasping for the Periphery: Ethnographic Imaginaries and Media Ecologies of Place in Late High-Growth Era Japan
No full textThis dissertation examines how media infrastructure reshaped popular perceptions of distant and peripheral sites in postwar Japan. It explores how media practitioners—across literature, film, television, and print—engaged with these regions to respond to emerging social, cultural, and ecological crises. Focusing on media production in historically “peripheralized” regions and former colonial sites during following the period of high economic growth, I trace how these spaces were materially transformed and mediated as sites of epistemological crisis and national reimagination. Through a close analysis of the “ethnography boom” of the 1970s, I examine how peripheral regions, former colonial holdings, and marginalized communities became focal points within Japan’s national imagination. I argue that these highly mediatized spaces—what I term “media ecologies of place”—played a critical role in shaping evolving notions of identity, space, and belonging. Utilizing a combined approach of structural analysis and close reading within the frameworks of media theory and cultural history, I explore how media creators experimented with ethnographic imaginaries across a wide range of intersecting—and at times oppositional—formats and genres. I contend that these media ecologies offer valuable insight into the emergence of new epistemologies of space, connectivity, national territory, and community. These frameworks continue to shape how “place” is experienced and imagined in contemporary Japan.
In each chapter of this project, I attend to the case study of a different kind of “peripheralized site” and core media genre to investigate how ethnographic and structuralist theories of cultural production were taken up by creators working across a variety of mediums to self-reflexively critique the structures of knowledge production at the heart of Japan’s modern consciousness. Chapter 1 examines the transformation of minicomi magazines and regional literature through the work of Kyushu-based writer Morisaki Kazue. Drawing on the structure of ethnographic works by canonical figures of modern Japanese thought, particularly Yanagita Kunio, Morisaki self-reflexively interrogated the intersections of language, cultural ideology, and modern consciousness. In tracing Morisaki’s shifting approach to documentary literature culminating in her 1973 work Gods of the Abyss: A Spiritual History of Coal Mine Labor, I consider how regional print cultures appropriated ethnographic frameworks to challenge dominant narratives of cultural homogeneity and historical continuity. Chapter 2 analyzes the documentary programs and theory of Konno Tsutomu, a co-founder of Japan’s first independent television production company, Terebi Man Union. Examining the breakout documentary travel series I Want to Go Far Away (1970-), I demonstrate how the concept of “place-ticity” (genbasei) emerged as a critical heuristic through which media practitioners responded to the creative and ethical demands of television production amid intense social upheaval, networking the spaces of the living room, the production site, and imagined ethnographic referents into new aleatory configurations. Chapter 3 evaluates the manga works of Mizuki Shigeru and Morohoshi Daijirō, tracing how their visual and narrative experimentation with yokai figures haunting the margins of the koma frame served as a means of creatively adapting and critiquing emerging sensibilities of “ethnographic realism” and the uncanny specters of colonial image culture that formed its roots. Chapter 4 investigates media events surrounding the 1972 reversion of Okinawa to Japanese administration and responses to these mediatic “calibrations” of cultural alterity. Using the concept of “(im)mobility” as a critical lens, I consider how the film Asia is One (1973) by Nihon Documentarist Union (NDU) and Okinawan playwright Chinen Seishin’s metatextual stage production The Human Pavilion (1976) negotiated the flows and constraints of ethnocultural subjects within an increasingly mediated nation-state, proposing spatial and temporal “counter-networks” to the models of alterity and belonging molded by the postwar state. I conclude with a brief discussion of the (re)institutionalization and operationalization of these ethnographic media imaginaries in the late 1970s, exemplified by the founding of The National Museum of Ethnology, or Minpaku, and the civilizational theories espoused by its first director-general, media theorist Umesao Tadao.East Asian Languages and Civilization
Neural Response Recovery Despite Persistent Cochlear Synaptopathy After Noise Exposure
No full textSynaptopathic noise exposure may preferentially target low-spontaneous rate auditory neurons. Here, we report on two models of noise-induced synaptopathy in gerbil in which hair cells remain intact and two that use higher-level exposures to produce outer hair cell (OHC) damage/loss in the cochlear base, where synaptopathy also is greatest. For all models, with and without hair cell damage, we consider a battery of functional measures that may help predict the extent and frequency location of synapse loss, as well as the nature of the fibers/synapses affected and their patterns of recovery. Experiments were conducted in Mongolian gerbil, an animal with a range of hearing sensitivity largely overlapping human, with known cochlear distributions of auditory nerve fibers by spontaneous rate (SR) subtype. Animals were noise-exposed then tested, with age-matched controls, at post-exposure time points from 24 hr to 36 wk. We recorded sound-evoked distortion product otoacoustic emissions (DPOAEs), compound action potentials (CAPs), and peri-stimulus time responses (PSTRs) across a broad range of frequencies, as well as unstimulated spontaneous neural activity. Hair cells and afferent synapses were quantified in immunolabeled cochlear tissue.
A single 2-hour octave-band noise exposure at 100 or 103 dB SPL yielded threshold shifts and response amplitude reductions (DPOAE, CAP) that recovered by 2wk post exposure, without hair cell loss. High-SR-dominated CAP and PSTR peak responses and spontaneous neural noise all recovered, exceeding control values at some post-exposure times. PSTR plateaus, reflecting contributions of neurons from all subgroups, also recovered but never exceeded controls. In the same ears, synapse loss was persistent, even 36 wk post exposure. With increased exposure level, synaptopathy was accompanied by permanent threshold shifts and persistent OHC injury or frank loss. After the 112 dB SPL exposure, permanent threshold shifts and mild OHC loss were restricted to the highest frequencies/cochlear regions evaluated. CAP amplitudes recovered more fully than DPOAE amplitudes. PSTR peaks and even plateaus were like control by 36 wk. Following the 115 dB SPL exposure, large threshold shifts and suprathreshold amplitude declines remained evident at the longest post exposure time. OHC loss was largest in the extreme base. For both higher-level exposures, spontaneous neural noise declined significantly, with incomplete post-noise recovery.
For a range of noise exposure levels, we observed persistent synaptopathy with recovery of neural response thresholds and amplitudes and augmented spontaneous activity. Through multiple functional assays and offline analyses, we focused on characterizing these neural responses. Post-noise recovery and even overshoot of control values for high-SR dominated onset responses was surprising given the synapse loss evident even in our longest held animals, perhaps signifying a compensatory mechanism. Our data suggest that there may be a range of noise doses that activates such a dynamic recovery process, whereas other exposures may be too low to activate it or too damaging to benefit from it.Medical Science
Explorations in Quantum Error Correction and Simulation of Topological Phases
No full textMuch excitement has arisen at the intersection between the field of quantum information theory with more traditional disciplines in physics such as high-energy and quantum condensed matter. This thesis consists of two independent explorations that are unified by the common language of quantum information: quantum error correction in holography and entanglement in topological phases of matter. The first part develops a fully algebraic formulation of quantum error correction tailored to the AdS/CFT correspondence. Building upon earlier work, it removes the assumption of a factorizable boundary Hilbert space, making the framework applicable to gauge theories such as super-symmetric Yang–Mills. This generalized framework yields a fully algebraic interpretation of the entanglement wedge reconstruction property in AdS/CFT as well as the Ryu–Takayanagi formula, and establishes an equivalence between the two. The second part of this thesis focuses on the exploration of interacting topological phases in strongly correlated lattice models. We begin by considering the setting of hardcore bosonic particles, and demonstrate that periodic driving naturally enables the realization of correlated hopping interactions that emulate flux attachment. Using large-scale density matrix renormalization group computations, we characterize the ground state of the resulting correlated hopping models on both the square and honeycomb lattice, finding bosonic integer and fractional quantum Hall states, respectively. Motivated by the possibility of adiabatically preparing such topological phases in cold atomic experiments, we map out the nearby phase diagram surrounding the bosonic integer quantum Hall state and propose an experimental implementation based upon laser-assisted tunneling of neutral atoms in a two-dimensional optical lattice. Finally, we turn to a classic, spin-model setting for exploring topological order: the nearest-neighbor spin-1/2 Heisenberg antiferromagnet on the Kagome lattice. We present a reexamination of the low-lying energy spectrum of this model using neural quantum states—a recently developed numerical method leveraging neural networks as a variational ansatz. While this method holds promise for simulating complex quantum systems, we highlight its limitations and potential pitfalls, emphasizing the importance of careful implementation and interpretation.Physic
Navigating Rank-and-file Opposition to Vietnam War: Case Study of the Vietnam War Debate, 1968
No full textHow did the rank and file oppose the Vietnam war within the labor movement in 1968, and how was the rank-and-file opposition to war contested? How did the Vietnam War debate reflect the distinctive challenges of contesting the leadership’s pro-war position within the Cold War labor movement?
To answer the research question, the study investigates the Vietnam war debate in the 1968 New York State American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) convention. Zooming in on the conversation between the anti-war rank and file and the pro-war labor unionists, the study identifies the reasons and strategies used by the anti-war rank and file and the pro-war labor unionists. On this basis, the study examines how the labor movement culture contributed to the difficulty of the anti-war labor arguments prevailing over the pro-war ones. The study identifies that the anti-war rank and file criticized the South Vietnamese government’s undemocratic behavior. The anti-war rank and file also criticized that the Vietnam war did not comply with the labor movement’s moral commitment to stand for the working people, arguing that the government funding for the Great Society programs decreased because of the war. A few anti-war rank and file directly showed their distaste for the drafting and claimed that no boys should go to war. On the other hand, the pro-war labor unionists argued that the anti-war rank-and-file arguments were driven by the false belief that they could make the Vietnam war decisions for the government. The pro-war labor unionists further argued that the labor opposition to war was an act of selfishness, cowardice, and disservice to the country. In addition, the pro-war labor unionists argued that fighting the Vietnam war was necessary because the war helped fight communism and defend democracy.
Comparing both sides of the arguments, the study concludes that the pro-war and anti-war labor primarily conflicted over their interpretation of democracy; identifies that some of the anti-war arguments were set aside because of their contradiction with the institutional culture. The study concludes that the pro-war labor position prevailed without solving the real concerns over the war’s impacts on the workers’ everyday life.
The study contributes to the research literature on the anti-Vietnam war social movement and the role of the Vietnam War in the development of the labor movement by studying how anti-war rank-and-file sentiments failed to translate into visible social
actions. Re-examining the interaction between the labor movement and the Vietnam war, the study finds that the rank-and-file anti-war sentiments were suppressed from within through a set of cultural norms of democracy, loyalty, and valor and contradicted by the labor movement leadership’s strong alliance with the government. In this process, the study exposes the connection between the Vietnam War debate and the tension beneath the image of the labor movement’s solidarity in relation to job type, gender, and race.Extension Studie
Thinking Outside the Black Box: Justifying Beliefs in the Age of Opaque Autonomous AI Systems
No full textIn March 2025, Manus AI emerged as “the world’s first fully autonomous AI agent” that’s capable of end-to-end decision-making without human intervention. Unlike previous AI systems, Manus represents a fundamental shift in AI-Human interactions as the system can independently identify problems, formulate approaches, execute solutions, and, most significantly, adapt its methodologies based on outcomes without human guidance. Yet this rush toward autonomous AI has neglected a fundamental philosophical question that should precede any deployment of self-governing AI systems: What is the epistemic status of beliefs formed based on outputs from AI systems whose operations remain fundamentally opaque to human understanding? When we deploy autonomous AI systems to make decisions, we are effectively authorizing them to form “beliefs” about the world and then act upon those beliefs.This question becomes particularly pressing in light of the “black box” nature of current AI systems. While these systems consistently produce accurate outputs, the processes through which they arrive at these conclusions remain fundamentally opaque. In this paper, I examine the epistemic opacity challenges of AI systems alongside an empirical analysis of current epistemic attitudes everyday AI users hold toward these systems, based on survey data collected on this subject. I will then argue that Goldman's reliabilist theory of justification offers a compelling solution by shifting focus to the reliability of belief-forming processes, which accommodates our counterintuitive acceptance of beliefs formed through processes whose operations remain opaque to us.Computer Scienc
The Revolution is in the Story: Where Community Voice Meets Strategy, Change Follows
No full textStorytelling is one of the oldest and most powerful ways we communicate. Compelling stories have the ability to unite people in social movements, transform beliefs, mindsets, and behavior, and inspire action. In education, storytelling holds immense potential to shed light on the meaningful work communities are doing to support children, youth, and families. Stories can bring the transformational power of this work to life—making what is complex feel achievable rather than overwhelming. Yet, far too often, these stories go untold because communities lack the time, capacity, resources, and support to communicate their impact in ways that influence public understanding, attract funding, change practices, and inform policy.
This capstone is grounded in the work I engaged in with The EdRedesign Lab at Harvard Graduate School of Education and its efforts to formalize how it builds storytelling capacity with its Institute for Success Planning communities. While these communities are driving real outcomes, many lack formal structures to document and share their stories of progress and impact. This results in missed opportunities to secure resources, advocate for systemic change, and counter deficit-based narratives.
In response to this issue, I developed a Storytelling Toolkit and Collective Impact Public Narrative framework (adapted from Marshall Ganz’s Public Narrative) to help communities craft stories uplifting their impact. The Storytelling Toolkit includes frameworks, tools, and resources that provide guidance on developing a storytelling strategy, using data to amplify stories, building trust with storytellers, elevating youth and family voices, and creating a storytelling culture.
My strategic project revealed that while tools and resources are essential, building a robust storytelling culture requires significant professional development, trust, coaching, and a shared belief in the power of stories. Communities need structured support to make storytelling a sustainable practice embedded in their organizational culture. Just as critical, EdRedesign must strengthen its internal capacity to support this work by continuing to model authentic storytelling, building capacity for staff, and ensuring community voice is elevated across all streams of work. By providing communities with a formalized structure to more effectively share their impact, EdRedesign helps them reclaim power—positioning them to define their narratives and share their impact on their own terms.Educatio
Trust-Based Algorithms for Resilient Multi-Agent Systems
No full textMulti-agent systems---ranging from robot teams to computer networks and distributed sensors---are increasingly deployed in real-world settings due to their advantages in scalability, parallelism, cost efficiency, and more. A core enabler of coordination in these systems is information exchange between agents, which can also become their Achilles' heel in the presence of malicious agents. Without any precautions, adversaries can severely degrade system performance simply by sending false data to other agents. Therefore, designing resilient algorithms capable of withstanding adversarial behavior is essential for enabling the safe deployment and widespread adoption of such systems, which is the ultimate goal of this thesis.
In this work, we consider settings where agents can leverage stochastic inter-agent trust signals, typically obtained from physical channels of information such as onboard sensors or fingerprints of wireless communication signals commonly available in cyber-physical systems. When used to quantify the trustworthiness of the transmitting agent or its transmitted data, prior work has shown that the availability of such signals can enable strong resilience guarantees---beyond what is achievable through data-based methods alone. Motivated by this potential, we focus on enhancing the resilience of distributed algorithms in multi-agent systems by incorporating inter-agent trust signals. To this end, we pursue three main directions: 1) developing detection and learning algorithms that identify legitimate and malicious agents beyond the applications and settings considered in existing work, 2) designing resilient distributed algorithms that integrate trust to solve fundamental multi-agent problems, including optimization, consensus, and hypothesis testing, and 3) deepening the theoretical understanding of resilience through trust observations by analyzing their limitations and studying how adversarial strategies affect the performance of the resulting algorithms.
We begin by addressing the joint challenges posed by directed communication graphs and the presence of adversaries. We develop a learning protocol that enables each agent to estimate the trustworthiness of all other agents in the system, extending prior detection results that were limited to learning about neighboring agents in undirected graphs. Next, we consider both constrained and unconstrained distributed optimization problems over directed graphs in adversarial settings, with a focus on achieving resilience while maintaining fast convergence. We first tackle the challenges introduced by directed graphs and constraint sets alone and propose a new gradient tracking-based algorithm, Projected Push-Pull, which achieves fast convergence under these conditions. We then extend this framework to incorporate adversaries by combining the learning protocol with Projected Push-Pull, resulting in an algorithm that achieves both fast and resilient convergence in the presence of adversarial agents. We investigate the performance of our algorithm in comparison to other resilient trust-based and data-only methods, and demonstrate that it outperforms them in various aspects, including convergence speed and the ability to eliminate the influence of malicious agents, while remaining applicable to a wider range of settings.
In the second part of the thesis, we shift our focus to challenges arising from the dynamic behavior of malicious agents, as well as the limitations of trust observations. To study these aspects, we consider the resilient consensus problem, where legitimate agents follow linear consensus updates while adversaries exhibit time-varying attack behavior, varying their probability of attack at each time step rather than attacking persistently as commonly assumed in the literature. We design a detection algorithm that uses a time-varying threshold: each agent first identifies its most trusted neighbors based on trust observations and then uses them as a reference to classify other agents. We characterize the performance of this detection algorithm, with particular attention to how it depends on the frequency of adversarial attacks and the choice of thresholds. Additionally, we analyze how dynamic adversarial behavior impacts consensus, specifically in terms of agents' ability to reach agreement and the deviation from the nominal outcome that would be achieved in the absence of adversaries.
Finally, we study two problems that highlight the impact of having a limited number of trust observations from each agent, as well as the influence of adversarial strategies on system performance. In the first problem, we revisit the resilient consensus setting, now considering a finite-horizon scenario where agents execute the consensus protocol for a known, limited number of steps. We model switching adversaries that behave cooperatively until a designated attack time, after which they send misleading data to maximize disagreement among legitimate agents. To address this, we propose a sliding-window-based detection algorithm that accounts for the adversaries’ switching behavior. We show that strong adversaries, who know when they will be detected, can compute their optimal attack strategy in polynomial time, and we provide bounds on the maximum disagreement they can induce among legitimate agents.
In the second problem, we study a binary hypothesis testing setting where a Fusion Center (FC) estimates the occurrence of an event after collecting binary measurements from distributed sensors. We consider a one-shot estimation scenario, in which the FC receives only a single measurement and a single trust observation from each sensor before making a decision. To address this, we develop the Adversarial Generalized Likelihood Ratio Test (A-GLRT), a polynomial-time algorithm that jointly estimates each agent’s trustworthiness and the adversarial strategy using both sensor data and trust observations, and then infers the true hypothesis based on these estimates. We deploy A-GLRT in hardware experiments emulating a crowdsensing application, where robots estimate and report traffic conditions on a mock-up road network under a Sybil attack, in which malicious agents spoof fake identities. By leveraging trust observations derived from the uniqueness of wireless signal characteristics, we show that A-GLRT outperforms existing resilient methods and maintains good performance even when malicious agents are in the majority.
In all parts of the thesis, we provide rigorous theoretical guarantees for the proposed algorithms, including results on convergence, convergence rates, and probabilistic bounds on detection and classification errors. Moreover, we support these findings with extensive numerical simulations that validate the effectiveness and applicability of our methods.Engineering and Applied Sciences - Computer Scienc