41656 research outputs found
Sort by
Three Essays in the Economics of Disagreements: Incentives, Measurement and Persistence
Get PDFThis dissertation presents three chapters that contribute to the study of economic forces shaping disagreements in society.
In Chapter 1, I demonstrate that political polarization can intensify due to innovations in the information market even if a population\u27s ideological distribution is fixed. Viewership-maximizing news firms cater to a diverse audience who assess source accuracy using noisy private signals that vary in precision and ideological bias. If better-informed consumers disproportionately migrate to newer platforms for news (e.g., the Internet), traditional media firms increase news slant to appeal more to less-informed partisans on both sides of the ideological spectrum. This leads to a greater divergence in beliefs about the state of the world among partisan traditional media audiences---increasing disagreements and potential hostility. The theory helps explain two empirical trends observed over recent decades: (i) rising slant in traditional news (e.g., cable television news) and (ii) increasing polarization among demographic groups least likely to use the Internet. I test the model’s central mechanism using an algorithmic text analysis of Facebook posts from 600 U.S. local television news stations. Media markets with greater expansions in moderately high-speed Internet access between 2012 and 2016 exhibit significantly larger increases in news slant, controlling for economic, demographic, and voting characteristics.
In Chapter 2, I present a method for measuring media slant using social media text data. Posts from political figures and media outlets are collected and processed through a large language model to extract “frames”---standardized declarative claims that encapsulate the core message of each post. These frames are embedded into a shared semantic space and clustered using a hierarchical density-based clustering algorithm, representing broadly held positions among politicians and media firms. For each account, vectors are constructed to represent their post frequency across each cluster. I separate the vectors for politicians and use party labels to train a classifier model that predicts political affiliation. The trained model is then applied to the vectors of media outlets, generating probabilistic scores that position each outlet on a partisan spectrum. This method contributes to the literature on measuring media slant using modern machine-learning tools to better discern semantic patterns in language and allows for granular separation of partisan positions and a nuanced measure of partisan slant without sacrificing interpretability.
In Chapter 3, I use a stylized compartmental model to analyze the long-term dynamics of misinformation propagation in social networks, focusing on the allocation of fact-checking resources. I conceptualize a false narrative as spreading through multiple types of claims, which can differ in their virality and resistance to fact-checking interventions. The analysis reveals that harder-to-debunk claims can persist when fact-checkers concentrate on easy-to-debunk claims—an approach commonly arising from crowd-sourced, consensus-based systems such as Community Notes—and ultimately become the primary vector sustaining the false narrative over time. I characterize the optimal allocation of fact-checking effort and show that, given sufficient resources, effective long-term mitigation of misinformation requires devoting resources to both easy and hard-to-debunk claims, no matter their initial virality or perceived cost. These findings challenge the prevailing focus on short-term fact-checking ``successes” and underscore the need to supplement crowd-sourced interventions with targeted professional fact-checking of complex or resilient misinformation. The theoretical framework provides actionable guidance for platforms and policymakers seeking to minimize the long-run societal impact of persistent false narratives
Unsupervised Deep Learning for Video Restoration
No full textIn today\u27s digital era, visual data is vital across several domains such as medical diagnostics, scientific imaging, surveillance, and entertainment. However, video data often suffers from degradations like noise, blur, compression artifacts, and low resolution, which degrade quality and downstream usability. Video restoration aims to recover clean, high-fidelity video from such corrupted inputs. Unlike static images, video restoration must maintain temporal consistency across frames, making it a significantly more complex problem. While supervised deep learning methods have achieved state-of-the-art results, they typically require large datasets of paired noisy-clean video datasets that are scarce or impractical to obtain in real-world settings like medical microscopy. Unsupervised methods overcome this challenge by learning directly from noisy data; however, current approaches are often limited in scope: they are often tailored for Gaussian noise, rely on large-scale datasets, or fail under low frame rates and complex motion. Moreover, most methods are tailored to a specific domain and struggle to generalize across diverse modalities such as fluorescence, confocal, and natural-scene videos. This work proposes a suite of robust unsupervised deep learning frameworks designed to overcome these limitations by handling multi-modal video data, adapting to varying motion dynamics, operating under limited data regimes, and effectively denoising across a broad range of noise types and intensities.
In the first study, we propose a novel unsupervised deep learning approach for video denoising that addresses data scarcity and demonstrates robustness against various noise patterns, enhancing its applicability. Our method consists of three modules: a Feature Generator that produces feature maps, a Denoise-Net that generates denoised yet slightly blurry reference frames, and a Refine-Net that restores high-frequency details. By employing a coordinate-based network, we simplify the architecture while effectively preserving fine details in denoised frames. Extensive experiments on simulated and real-world videos, including calcium imaging, show that this approach can denoise corrupted videos without requiring prior noise model knowledge or extensive data augmentation during training.
In the second study, we extend our work to medical imaging by introducing a Deep Temporal Interpolation method, which leverages temporal correlation across long video sequences. This method incorporates a temporal signal filter into the lower CNN layers to restore microscopy videos corrupted by unknown noise types. Our unsupervised framework adapts to diverse noise conditions without requiring prior knowledge of noise distributions, addressing a critical gap in real-world medical applications. Evaluations on real microscopy recordings and simulated data confirm our framework\u27s superior performance across a wide range of noise scenarios. Experiments show that our model consistently outperforms state-of-the-art supervised and unsupervised techniques, making it effective for denoising microscopy videos.
To generalize across the various multi-modal microscopy imaging including 2D time-lapse sequences of different temporal rates and 3D stack volumetric images while maintaining its efficacy in a limited training data setting, while remaining effective in limited‐data settings, the third study presents an innovative Spatio-Temporal Sampling method that generalizes effectively across diverse microscopy imaging modalities without requiring per-modality tuning. Our network integrates a learnable Spatio-Temporal Weighted (STW) kernel guided by optical flow estimation to balance temporal information, a Guided Temporal Fusion module that adaptively warps and gates high-frequency details from the target frames, and a recurrent optimization loop for iterative refinement of optical flow and denoised outputs. Extensive experiments demonstrate that our model effectively denoises microscopy videos, maintaining structural detail and temporal coherence, with multi-modal generalization ability.
In the fourth study, we introduce VR-INR, a novel video restoration method leveraging Implicit Neural Representations (INRs). Our approach integrates hierarchical spatial-temporal-texture encoding with multi-resolution implicit hash encoding, enabling adaptive reconstruction of sharp, noise-free frames from low-resolution inputs. VR-INR is trained solely at a single scale (×4) and effectively generalizes to arbitrary super-resolution scale factors at test time. Furthermore, our model demonstrates zero-shot denoising capabilities without prior training on noisy data. Extensive experiments validate VR-INR\u27s superior performance over state-of-the-art methods in sharpness, detail preservation, and denoising across diverse unseen scales and degradations
Item 2: Activity 1: Creating Group Agreements for Classroom Discussions
No full textHigher education classrooms are ideal for learning the skills of democratic civic discourse, but these spaces also need agreed upon norms for how students will engage each other. Over four sequential steps, this activity asks students to reflect on the times they’ve had a difficult discussion, what went well and what did not, and to use these reflections to generate a set of shared class values to guide their conversations about wicked issues
Item 2: Activity 1: African American Literacy, Literature, and Voting Rights
No full textThis curricular unit, deliverable in person or online, seeks to introduce students to the arc of voting rights in and adjacent to African American literature, including the practices of African American literacy in claiming and exercising the right to vote, from Reconstruction to the Civil Rights Movement. G. Ellis Harris’s North Carolina Constitutional Reader (1903) represents a textual artifact of African American agitation for literacy and thus the exercise of the right to vote following the termination of Reconstruction, and Alice Walker’s Meridian (1976), contextualized by Orlando Bagwell et al.’s Eyes on the Prize: America’s Civil Rights Movement (1987), roots students in their own coming to consciousness and subsequent civic engagement. Culminating in a presentation by a state board of elections official supplemented by a quiz assessing understanding of current election law, the activity seeks to empower students with historical context, underscore the power of literacy and literature to cultivate civic consciousness, and prepare students to exercise their right to vote
Item 2 and 2.1: Activity 1: Who\u27s Interested in the News? Whose Interests Are in the News?
No full textThis activity requires students to compare and contrast characteristics of popular or commercial news stories with public interest news stories. Students will make critical distinctions between different kinds of information that coexist under the broad umbrella of “news” in a contemporary information ecosystem. Instructors should provide students with the worksheet for this activity
Interactions of Biorational and Synthetic DMI Fungicides for Control of Fruit Diseases
Get PDFBiological fungicides are increasingly investigated as sustainable alternatives to synthetic fungicides in fruit crop protection. This dissertation examines the interactions between two commercial biological fungicides, Bacillus subtilis AFS032321 (Theia) and metabolites of Pseudomonas chlororaphis AFS009 (Howler EVO); and demethylation inhibitor (DMI) fungicides, with a focus on compatibility, synergistic potential, and resistance implications. In vitro assays revealed that commonly used DMI fungicides difenoconazole, metconazole, tebuconazole, and propiconazole inhibited vegetative growth of B. subtilis at concentrations above 50 µg/ml, whereas mefentrifluconazole promoted colony expansion. Under controlled conditions, lesion development on detached fruit was inhibited synergistically when Theia was combined with the growth-promoting DMI mefentrifluconazole, while antagonistic effects occurred with inhibitory DMIs like difenoconazole. Fungicide formulations also influenced compatibility: non-emulsifiable concentrate (non-EC) formulations of propiconazole and metconazole were less inhibitory to B. subtilis and resulted in smaller lesions on detached apples and cherries when used in combination with Theia. The product reduced blossom blight in nectarines and peaches, suggesting its potential as an alternative to chemicals for early-season sprays.
Synergistic interactions between DMI fungicide propiconazole and Howler EVO were observed and reduced rates of the DMI mixed with Howler EVO were as effective as the full rate of the synthetic fungicides. Investigations into modes of action revealed that Howler EVO suppressed constitutive and DMI-induced expression of the MfCYP51 gene in DMI-sensitive and -resistant Monilinia fructicola isolates. This was later confirmed to be caused by the presence of pyrrolnitrin (PRN), a secondary metabolite produced by P. chlororaphis and present in Howler EVO. Field trials confirmed that Howler EVO combined with a reduced rate of propiconazole reduced blossom blight and brown rot in nectarines. Our findings also established that at least some biologicals should be used in resistance management programs. Under controlled conditions, Botrytis cinerea isolates resistant to the synthetic fungicide fludioxonil were also resistant to biological fungicide (Howler EVO). This cross resistance was likely due to the chemical similarity between active ingredient PRN in Howler EVO and fludioxonil. Overall, this work highlights both the promise and the complexity of integrating biological fungicides into modern disease management programs and provides critical insights into optimizing their use for sustainable fruit crop production
Human Comfort Modeling, Measurement, and Improvement in Human–Robot Collaboration
No full textA dissertation is proposed to explore human comfort in human-robot collaboration (HRC) through modeling, prediction, and enhancement methodologies. Human comfort is a crucial yet underexplored factor in HRC, directly influencing task efficiency, trust, and overall collaboration effectiveness. Understanding the influential factors, developing computational models, and refining methods to improve human comfort in HRC are essential steps toward advancing the field of collaborative robotics. To address these challenges, multiple studies have been conducted. A series of experimental studies were performed to investigate how robot motion-based parameters affect human comfort in HRC. These studies examined both analytical comfort modeling approaches and physiological signal-based detection techniques, offering insights into how robot behaviors impact human comfort. In addition, comfort-aware robot behavior control strategies were designed and validated, demonstrating the feasibility of dynamically adapting robot actions to enhance user comfort. Moreover, comparative analyses between reality-based and VR-based HRC environments were carried out to explore the feasibility of using immersive technologies for comfort evaluation and adaptation. Building upon these prior efforts, the proposed research seeks to further enhance human comfort in HRC through three main directions: (1) developing a comfort-centric task allocation framework using a Markov Decision Process (MDP) to generalize and extend comfort modeling to more complex real-world collaboration tasks; (2) integrating large language models (LLMs) for adaptive planning, utilizing an RLHF-inspired approach to align task planning strategies with human comfort preferences; and (3) enhancing robot intention communication through augmented reality (AR), ensuring that humans can intuitively perceive robot intentions and actions, thereby reducing uncertainty and improving overall interaction quality
Item 7: Syllabus
Get PDFThis course will invite students to engage in conversations about connecting (and avoiding), helping and learning in our civil society. We’ll think and talk about thorny questions of what it means to be a citizen of ASU, the High Country, the US, and the world. Through the reading of provocative texts and a few films, we’ll practice the habit of civil discourse and respectful argumentation as we engage in service activities. Basically, we’ll practice the habit of talking in community about our relationship to community. Students will participate in community-based service-learning to explore the themes of civic connecting, helping and learning (especially via talking). The course will improve the student’s ability to analyze, evaluate, and synthesize complex ideas in the context of current community issues.
The First Year Seminar (UCO 1200) provides students with an introduction to the four goals of a liberal education at Appalachian State University. Specifically, students will practice (1) thinking critically and creatively and (2) communicating effectively. In addition, students will be introduced to the learning goals of (3) making local-to-global connections and (4) understanding responsibilities of community membership. While each First Year Seminar course engages a unique topic examined from multiple perspectives, each course also introduces students to a common set of transferable skills. As such, this First Year Seminar facilitates student engagement with fellow students, the University, the community, and the common reading as well as essential college-level research and information literacy skills and the habits of rigorous study, intellectual growth, and lifelong learning
Item 5: Assignment 2: The Social Media Posts of Puerto Rico’s Main Political Parties
Get PDFThe purpose of this assignment is for students to analyze the differences among Puerto Rico’s main political parties, and how they represent themselves on social media. Students work in small groups to conduct content analysis of the Facebook posts of Puerto Rico’s main five political parties. After reviewing a brief description and history of each political party, students access their Facebook accounts and review the last twenty-five posts that each party made. Then, using a guide included in this assignment, they analyze the content of those posts to determine: the main topics addressed through social media; the sentiment attached to these themes; and reactions from the public
Item 4: Activity 1: The Relational Syllabus GPT – A Faculty Tool for Inclusive Civic
No full textThis faculty-facing GPT tool supports inclusive, equity-based course design by helping instructors create assignments and syllabi that engage students in civic learning, democratic reasoning, and ethical AI use. Drawing from Universal Design for Learning (UDL), Transparency in Learning and Teaching (TILT), and Action-Oriented Competency Didactics (AOCD), the GPT scaffolds syllabus revision, integration of civic learning outcomes, and multimodal assignment redesign. The tool includes templates for AI use and reflection, accessible tone adjustments, and civic-based learning progression. It is designed for faculty across general education, political science, and interdisciplinary humanities, and supports both AI-assisted and AI-free learning paths