Nations censor specific information in accordance with their political, legal, and cultural standards. Each country adopts unique approaches and regulations for censorship, whether it involves moderating online content or prohibiting protests. This paper seeks to study the underlying motivations for the disparate behaviors exhibited by authorities and individuals. To achieve this, we develop a mathematical model designed to understand the dynamics between authority figures and individuals, analyzing their behaviors under various conditions. We argue that individuals essentially act in three phases - compliance, self-censoring, and defiance when faced with different situations under their own desires and the authority's parameters. We substantiate our findings by conducting different simulations on the model and visualizing the outcomes. Through these simulations, we realize why individuals exhibit behaviors falling into one of three categories, who are influenced by factors such as the level of surveillance imposed by the authority, the severity of punishments, the tolerance for dissent, or the individuals' boldness. This also helped us to understand why certain populations in a country exhibit defiance, self-censoring behavior, or compliance as they interact with each other and behave under specific conditions within a small network world.

Speech to text models have become a very useful tool for hospitals. Hospitals can use automatic transcriptions to be able to reduce workload on doctors and clinicians since they do not have to manually record information. This automation can give them more time to meet with more patients and increase the efficiency of hospital work. However, an unexplored application of speech-to-text are emergency calls. The most common use for automated transcriptions are to document what doctors are doing and are given time to proofread for errors. This work focuses on the problem of transcriptions of emergency call data. Our work curates this emergency call data and models it as a medical transcription problem in hopes that the transcriptions can be used later for medical decision making. The heavy background noise and poor audio quality that comes with emergency radio are the reason this problem is challenging to solve. The results of this experiment show a modest increase to the accuracy of transcribing the emergency hospital recordings.

For this creative project, I created a visually immersive and artistic data visualization of global space-related activities. The project aims to create a sense of wonder and creativity for space exploration through unconventional data visualization. By focusing mainly on the artistic elements of the visualization, the project will have a larger emotional impact on its viewers, as opposed to a traditional data visualization. The project uses a comprehensive dataset of space-related articles, all of which include the location of the activity discussed in the article, as well as keywords and other fields. The dataset will serve as material to create a narrative that shows not only how space-related activities are distributed around the globe but also the overarching themes of the activities. To create the final project, I used the JavaScript library p5.js.

The scientific manuscript review stage is a key part of the modern scientific process. It involves rigorous evaluation of new papers by peers to assess the significance of contributions in a particular area of study and ensure that papers meet high standards. This process helps maintain the quality and credibility of research. However, some reviews can be toxic or overly discouraging, leading to unintentional psychological damage (such as anxiety or depression) to paper authors and detracting from the constructive tone of the review space. This Thesis/Creative Project was completed alongside a capstone project. Our capstone project aims to address this issue. The goal is to fine tune a Large Language Model (LLM) that can first accurately identify toxic sentences within a paper review. Then, the LLM will revise any toxic sentences in a way that maintains the criticism but delivers it in a more friendly or encouraging tone. To effectively use this LLM, it requires a Graphical User Interface (GUI) so that end-users (such as editors, associate editors, reviewers) can easily interact with it. This allows them to update the wording of the review in an effective manner while maintaining scientific integrity. While the GUI provides a user-friendly interface for interacting with the LLM, there are some technical challenges in running a LLM application in a web-based framework. LLMs are computationally expensive to run. They require significant GPU RAM, which can be a limiting factor, especially in a web-based framework with limited resources. One potential solution to this problem is model quantization, which can reduce the memory footprint of the model. However, this introduces the problem of model drift, as the model’s performance may decrease when quantized. This needs to be measured to ensure the model continues to provide accurate results.

Many companies wish to create a personalized experience for customers using their websites. For many services that might mean changing the icon on a sign in page; however, as computers become more powerful and customers expect more from the services they use, companies are starting to investigate ways of running personalized code for their customers. Sadly, one big problem with this trend is that it is very new. This leads to many problems such as the lack of technologies fit for a certain scenario, the flooding of new technologies in only a specific field, and the overall general confusion in implementing a novel technology like this. This is why I believe that compounding a list of different technologies, each with a list of pros and cons, example implementations to give a feel of the technology, as well as benchmarks of each method to allow for individuals and companies to create better websites and services for their customers. I will also be going through a history of available technologies to give an idea on how this technology used to be used for, how it is used today, and how I believe it will be used in the future.

Machine learning algorithms have a wide variety of applications and use cases. They are robust in the sense that they can continue to learn and improve long after they have been deployed without much programmer supervision. One key area that machine learning has been used for is in the detection and classification of objects in images and videos. This so-called computer vision has typically been used by companies to extract user information from the images and videos that they post. Meta (formerly known as Facebook) had been using such algorithms to automatically tag users in pictures that were uploaded to the Facebook website up until November 2021 [1]. Although these algorithms have been used to exploit user’s privacy, they can also be used to help ensure this privacy. For this creative project, I developed a machine learning model that could detect faces in a given picture and identify the area of the picture that these faces took up. Training a model from scratch can take millions of images of data and hundreds of hours on powerful GPUs. Since I didn’t have access to those resources, I began with a pre-trained model known as VGG16 by Karen Simonyan & Andrew Zisserman. From there, I took 90 pictures of myself and annotated where in the image my face was located. Since 90 pictures wouldn’t be enough data for this algorithm, I used an image augmentation algorithm to randomly crop, flip, change brightness, change gamma, and recolor the images to expand the dataset. In total, I used 5400 images to train the algorithm. The machine learning model had a loss value that hovered around 0.1 thanks to the VGG16 model. It was able to accurately detect my face and also adapt whenever I moved my face horizontally and vertically across a camera. However, the model struggled to draw a bounding box whenever I moved my face forward or backward in the camera shot.