116964 research outputs found
Sort by
The Climate of AI: Profit of Planet?
Get PDFArtificial intelligence has become a cornerstone of technological progress, yet its environmental and social implications are insufficiently examined. Data center demand grows exponentially while innovation outpaces regulation. While research on climate change and on AI exist independently, there is a notable gap in scholarship on AI’s climate impacts. The overlap does exist but is fragmented and lacks public accessibility.
This thesis primarily asks: how do we bridge the gap between technological advancement and environmental consciousness? To answer this, I examine AI’s environmental impacts, sustainability of current technical and government structures, and how to keep innovation sustainable. This thesis is timely; as AI dependence increases, energy and climate demands rise. All while U.S. regulatory protections weaken.
I take a mixed-methods approach—including a meta-literature review, perspective mapping, framework creation, policy, and budget analysis. I argue that AI’s environmental harms are systemic and rooted in late-stage capitalism. My research identifies five key problem dimensions: resource extraction, emissions and energy intensity, transparency deficits, economic externalities, and ethical impacts. After analyzing these and comparing them to historical legislation, I provide actionable solutions via a framework. Then, I meticulously validate the concepts by analyzing the EU AI Act. This thesis argues that mandatory environmental reporting, efficiency standards, renewable-aligned incentives, and enforceable regulatory constraints are essential to achieving sustainable AI.
My research organizes every major environmental, social, and health argument relating to AI, and makes them accessible to a general audience. This thesis provides an actionable framework that can be used by policymakers, companies, and individuals to understand and create positive differences.Plan II Honors Progra
Global change biology at population margins
Get PDFGlobal ecosystems are changing at an unprecedented rate due to human activities. These changes pose a huge threat to the biodiversity and ecosystem services they provide, directly affecting human well-being. Therefore, a fundamental challenge in modern biogeography is to develop a theoretical framework to understand how global change shapes biodiversity patterns. A natural question to ask is—what should be the conceptual basis of the theory of global change biology? I argue that the theory of global change should be embedded within the framework of population demography. After all, population change patterns result from four fundamental demographic processes in population biology—birth, death, immigration, and emigration. In my thesis, I take this demographic approach to address three standing problems in global change biology at population margins. I study population margins because many global change phenomena, such as range shifts, the spread of invasive species, and persistence in fragmented landscapes, are spatial. And in contrast to the population core, margins are subjected to a unique set of ecological and evolutionary pressures. Thus, understanding population dynamics under these novel conditions might allow us to discover new biological mechanisms that regulate population change. In the first chapter, I show that source-sink dynamics could be as important as the climate in determining species ranges at large-spatial scales. In the second chapter, I propose an age-structured metapopulation model to describe the spread patterns of zebra mussels via commercial shipping networks in inland North America. And finally, in my third chapter, I examine how non-random recolonization after local extinctions due to extreme events can lead to the rapid evolution of dispersal traits in soapberry bug metapopulation. Overall, my dissertation highlights how population demography provides a unified perspective to study global change biology and provides novel ecological and evolutionary insights on how diversity is maintained at population margins.Ecology, Evolution and Behavio
Design and behavior of seamless bridge-pavement systems
Get PDFThe seamless bridge concept eliminates expansion joints on the bridge deck and joints between the bridge and approach, which can significantly reduce the maintenance costs and improve the long-term durability of the primary load-carrying components. Past applications of seamless bridges have utilized Continuously Reinforced Concrete Pavement (CRCP) in which a transition zone is employed between the bridge deck and CRCP to accommodate deformations caused by the longitudinal expansion and contraction of the bridge. A critical aspect of the system response is the longitudinal load transfer mechanism in the transition zone, which is governed by the concrete slab-base interaction. This dissertation aims to advance the development and implementation of seamless systems in the U.S. based on a comprehensive investigation including experimental testing and numerical modeling. The experimental study focused on the characterization of the concrete slab-base interaction through unit-cell direct shear tests and cyclic large-scale push-off tests. The load (shear) versus displacement behavior at the concrete slab-base interface was evaluated for different interface materials (geotextiles, polyethylene sheets, felt paper) intended to break the bond and control the level of interface restraint. The effects of cyclic movements were investigated. Test results indicated that polyethylene sheets and felt paper were effective in eliminating the bond between concrete and base. The numerical study involved the development of a structural model of the entire seamless system under longitudinal effects, and a more detailed continuum finite element model of the transition zone under combined longitudinal and out-of-plane (vertical) effects. The numerical models were used to study the influence of the friction coefficient at the concrete slab-base interface, reinforcement ratio and slab thickness of the transition zone on the axial response of the system. Potential design issues related to the seamless bridge connection were investigated through numerical parametric studies for various bridge prototype structures in Texas with different configurations. Based on the experimental results and numerical studies, double-sided textured linear low-density polyethylene sheets and felt paper are the most promising interface materials to be considered for the transition zone of a seamless bridge-pavement system. Finally, a design procedure with general recommendations for the seamless bridge-pavement systems is proposed.Civil, Architectural, and Environmental Engineerin
The marvelous feminine : a queer deconstruction of femininity, feminism, and commodity in the Marvel Cinematic Universe
Get PDFHow does the superheroine create slippages between traditional femininity, queerness, and cisgendered womanhood? How are these slippages presented when the heroine functions as a feminist commodity within cis heteropatriarchal capitalism? This thesis aims to articulate the role of marketable liberal feminism as a conduit for an oppositional encoding of queerness and transfemininity via the Marvel Cinematic Universe and the characters of Captain Marvel and She-Hulk. Section I surveys feminist narratives and characteristics of both She-Hulk and Captain Marvel, articulating the refusals to adhere to hegemonic, cisheteropatriarchy are encoded as liberal feminist rhetorics for commercial interests. In Section II, I discuss the previously proposed liberal feminist rhetorics through an oppositional encoding of queer codification and a lesbian gaze exhibited by the character of Carol Danvers aka Captain Marvel in the film Captain Marvel (2019). In Section III, I explore the character of Jennifer Walters aka She-Hulk in the Disney Plus series She Hulk: Attorney at Law (2022) and examine liberal feminist rhetorics of the body, femininity, and self-determination as transfeminine oppositional encoding. I conclude with future areas for exploration in circumscribing queerness into the subtextual surface of mainstream media entities.Women's and Gender Studie
Unveiling the public park system in the Distrito Central, Tegucigalpa : assessing park development and community impacts
Get PDFThis study explores the program's structure, park development processes, and the tangible impacts of the selected parks from Fundacion Vamos al Parque (FVP) have on their surrounding communities. Through a blend of qualitative methods, including user observations, questionnaires, and site analysis, this research scrutinizes both the procedural intricacies of park construction within the Vamos al Parque program and the localized effects of these green spaces on community dynamics, well-being, and social integration. The investigation is initiated by unraveling the organizational framework and procedural intricacies inherent in park establishment under the Vamos al Parque program. By assessing park quality from users' perspectives and scrutinizing the policies and design strategies employed in park development, this research seeks to comprehend the program's structure and processes and illuminate the tangible effects of park investment on community empowerment and cohesion. Adopting a holistic approach to park design and management can prioritize the diverse needs of the community, promoting the fundamental right of citizens to public parks and changing the current perception of parks as just a means of promoting safety. Finally, this research makes a significant contribution to the ongoing conversation on urban development and community engagement by offering valuable insights into the efficacy of park development strategies.Community and Regional PlanningArchitectur
Hierarchical and compositional reinforcement learning for autonomous systems
Get PDFWe explore the use of hierarchy and composition for developing reinforcement learning-based controllers for the navigation of autonomous agents in complex settings. This thesis consists of three segments: In the first segment, we develop a hierarchical control system for real-time head-to-head racing, employing a discrete game formulation for the high-level planner with Monte Carlo Tree Search (MCTS). The solution generates long-term waypoints that adhere to the rules and approximate a Nash equilibrium. A streamlined variant of the general racing game is created to follow waypoints and generate high-resolution control inputs through multi-agent reinforcement learning. This hierarchical control scheme surpasses previous autonomous racing methods in performance and safety rule adherence, exhibiting behavior akin to skilled human drivers. In the second segment, we present a framework for verifiable and compositional reinforcement learning (RL), consisting of a high-level model and low-level subsystems. The high-level model, depicted as a parametric Markov decision process, facilitates the planning and analysis of subsystem compositions. Interfaces between the subsystems enable automated decomposition of task specifications, allowing independent training, evaluation, and repurposing of the subsystems for task transfer situations. In the third segment, we demonstrate a proof-of-concept for a hierarchical framework that leverages compositionality to execute multi-agent cooperative tasks in a warehouse robot navigation and delivery setting. Employing a modified, variable-step single-player version of Monte Carlo Tree Search (MCTS) as the high-level controller, a joint policy for all agents is formulated by simulating an abstract state and action space. The joint policy determines the sequence of subcontrollers for each agent, maximizing efficiency and performance. This proof-of-concept showcases the potential of our hierarchical and compositional framework in solving complex multi-agent cooperative tasks in various real-world applications.Aerospace Engineerin
Task-aware planning and learning in partially observable environments
Get PDFAutonomous systems are increasingly being deployed in the real world. The decision-making agents in these systems often rely on sensors that may only provide them with an incomplete picture of the world. Such partial observability manifests scalability issues that are not present when the decision-making agent has complete information. The designers of autonomous systems are more frequently applying techniques from artificial intelligence to address these issues. Yet, as these systems become tightly integrated into society, there is a requirement to ensure that they are safe. We present theory and algorithms for decision-making agents making verifiably correct decisions under incomplete information. In particular, we provide a method that integrates scalable recurrent neural network representations from machine learning with the provable guarantees of formal verification. Additionally, we present two approaches that ensure the agent will make correct decisions by restricting unsafe choices. One employs a formally-verified shield, constructed from a partial system model, to ensure that an agent performing reinforcement learning in a partially observable environment will remain safe both during and after learning. The other approach applies a factored state-representation to reason about the possible decisions of an adversarial agent. Finally, we devise a systematic approach for a team of mobile agents performing the task of decentralized classification in partially observable environments. By combining resilient information sharing algorithms with assurances from contract-based synthesis, we provably guarantee that the team will converge on the correct classification in the presence of adversaries.Aerospace Engineerin
Master's thesis recital (oboe)
No full textPoem for oboe and piano / Marina Dranishnikova -- Seduction dance : for flute, oboe and piano / Miguel del Aguila -- Fantasia sull' opera Poliuto di Donizetti : for oboe and piano / Antonio Pasculli -- Oboe quartet in F major, K. 370 / W.A. Mozart.MusicName of supervisor not provide
‘How much do you think that hurt me?’ : age, race, and sex biases in pain assessments conducted by adults and children
Get PDFPain treatment inequalities are linked to common biases about demographic influences on pain sensitivity, including effects of age, race, and sex. The physical pain of women, children, and Black individuals is undertreated compared to that of of men, adults, and white individuals (Brennan-Hunter, 2001; Dao & LeResche, 2000; Wandner et al., 2012). Past research has examined these biases individually, but the present research examines all three together. In the present studies, participating adults (n=307) and children (n=71, age 4-5 years) read vignettes about characters suffering injuries, and rated the characters’ pain on a 1-7 pain scale. These vignettes combined specific injuries (e.g., papercut; broken arm) and photos of character faces representing diverse age, race, and sex demographics. We also collected participant age, race, and sex information. Mixed effect linear regression models were used to analyze the data for effects of character and participant demographics. In Study 1 we found that two of our three expected biases toward the characters were significant in adult participants. Adult participants rated child characters as more pain-sensitive than adult characters, and female characters as more pain-sensitive than male characters. Adult participants also demonstrated an interaction of age and sex in which adult female characters were rated as more pain sensitive than adult male characters, yet male and female child characters were rated comparably. Unexpectedly, adult participants did not demonstrate a character race bias. However, adult participants’ race had an effect: white participants gave the lowest pain ratings, Black participants gave the highest pain ratings, and Hispanic participants’ rating average was between white and Black participants. Patterns in 4- to 5-year-old children’s ratings in Study 2 revealed they are still learning to differentiate specific injuries and how painful they are. In contrast to results with adult participants, child participants' pain ratings exhibited no biases. Two explanations are possible. One, these biases emerge after the age of 5, either later in childhood or adulthood, suggesting that substantial cultural exposure is required to form them. Alternatively, task demands with the scale, specific injury experience, and perspective taking demands may be too high for 4 to 5-year-olds, obscuring their biases. Future research should include more adult participants from the US, and either child sample of older children, or a simpler rating scale.Psycholog
A Bayesian longitudinal network analysis of panic disorder symptoms and respiratory biomarkers
Get PDFThe network theory of psychopathology is gaining popularity as a conceptualization of psychological disorders that may aid the identification of mechanisms of therapeutic change. However, many existing networks do not consider other, relevant variables beyond the symptoms themselves. We present a large-scale (n = 1,873), longitudinal Bayesian network analysis of panic disorder using the symptom items from the Panic Disorder Severity Scale (PDSS) and two respiratory biomarkers (respiration rate and end-tidal CO₂) collected during routine monitoring of a Capnometry Guided Respiratory Intervention (CGRI). Our findings offer support for avoidance and fear of panic as drivers of subsequent panic disorder symptoms over the four-week course of treatment. Moreover, respiration rate but not end-tidal CO₂ was associated with downstream PDSS symptoms. These findings provide further evidence supporting the role of respiratory biomarkers in the maintenance of panic disorder and some support for normalization of dysfunctional breathing as one therapeutic mechanism governing CGRI.Psycholog