Lehigh AI

The most common sources of user errors with data science models and how to prevent them

This project outlines the most common sources of errors users face with data science models, which includes when a model returns wrong outputs, varies output by release, crashes, misses the latest features, and doesn't run in a new environment. Solutions to these issues are demonstrated with coding examples.

This research explores how writing teachers' and multilingual students' use of AI can shape their perceptions of ethics in composition.

Multilingual English students often encounter difficulty writing research in English. In their first year, they must develop sufficient proficiency to attribute words and ideas to other sources, while learning the culturally defined ethical binaries regarding originality vs. plagiarism. A coping strategy for them is to use AI tools to explore how to write, and how to use feedback to revise. But where is the boundary between using AI to promote vs. replace learning? And how does using AI shape students' perceptions of their own writing processes?

As a means to "demystify" generative models, I will develop a course module that begins with basic (first generation) language models to create simple autocomplete processes, and then move on to more complex implementations, such as RNNs, LSTMs, and transformers. This module will use R and Python, as well as open-source tools like Ollama.

While GUIs for generative models and integration into existing apps is proliferating, many

Combines Ai, active study techniques to help students get better grades in less time.

Study companion will be a vertical llm agent built upon chat gpt to be the most comprehensive study tool. A study tool that tell you when to study in your day and for how long, what content to study to be on schedule to meet your goals and how to study it with the most effective techniques, in order to get better grades in less time.

Develop a LLM model evaluation guide, framework, and workshop to promote AI and prompt engineering literacy at Lehigh University.

The rapidly evolving AI landscape presents continuous advancements in performance, features, and functionality. Vendors frequently release new updates, often outpacing each other within a
span of months. For instance, even within a single vendor like OpenAI, the release cadence is challenging to follow.

Fine-Tuning Large Language Models for Enhanced Research Outcomes and Insights with Domain-Specific Dialogue

Large Language Models (LLMs) have become pivotal technology, enabling machines to understand and generate human-like replies to questions. Popular pre-trained LLMs have a general understanding of language and capture a wide range of linguistic patterns, but may not perform in specific tasks or domains. Specialized training or fine-tuning is needed to help improve their performance and accuracy.

Students engage in a multi-week sales project, using Generative AI to develop sales strategies while considering ethical implications.

Challenge (or Opportunity) // Many universities don't specifically teach the sales process. However, creating and applying innovative sales strategies can enhance an emerging entrepreneur's confidence, reach more segments of the target market, and speed up organic growth of the company. Sales skills are often best acquired through hands-on experience, involving trial and error, reflection, and adaptation.

To better understand and quantify the impact of our research efforts, we propose a generative AI use case project aimed at creating a comprehensive prompt to measure research impact. This project will explore various dimensions of research impact, ensuring a holistic approach to the evaluation.

The primary objective of this project is to develop a robust generative AI prompt that can effectively draw out and/or provide a measurement of the research impact within the College of Health. This prompt will serve as a tool for Research Administration staff to help faculty researchers assess the contributions of their work in terms of academic influence, societal benefit, and policy implications, among other measurements of Research Impact.

The goal of this assignment is to produce student generated epidemic models that can address a healthcare issue using—entirely—ChatGPT.

There is an immediate need for public health practitioners to incorporate mathematical modeling into infectious diseases decision making. This is demonstrated by a continued increase in the use of forecasting models by state and federal health agencies such as the CDC. However, the typical public health curriculum does not require mathematical modeling and programming of epidemic models. In my class, Outbreak Science and Public Health forecasting, one week will be devoted to the use of Large Language Models to construct epidemic models that can incorporate temporal public health decisions.

Systematic review of AI ethics education in K-12 and higher education to uncover themes, challenges, and teaching practices.

This project focuses on conducting a systematic review of AI ethics education for K-12 and higher education, addressing the critical gap in understanding the current state of teaching AI ethics. With the rapid integration of AI into daily life, there is an increasing need to ensure that students not only grasp fundamental AI concepts but also develop the ethical awareness necessary to responsibly use and create AI technologies. This project how and what AI ethics topics have been introduced in educational contexts. Currently, I am in the data extraction stage of the project.

Integrating AI-powered tools in diagnostics and treatment of Speech Sound Disorders, improving accessibility and reducing Speech-Language Pathologist workload.

I utilized AI-powered speech analysis tools into speech therapy to enhance the diagnosis of Speech Sound Disorders (SSD). Given the limited access to Speech-Language Pathologists (SLPs) and the large number of people with the need for Speech Therapy in the United States, there is a large caseload for SLP's in America. My work aims to bridge this gap by utilizing AI technologies such as automatic speech recognition and large language models.

This lesson template provides a non-technical introduction to ethical considerations about AI, through the lens of a Captain America film.

We are bombarded with claims about the ways that AI algorithms are remaking society, often made with hyperbolic comparisons to science fiction. How should we evaluate these claims? Though they hear about the impact of "algorithms" all the time, students often do not know what they are or how they work, unless they are pursuing courses in computer science. In order to effectively evaluate the news around them, and understand which tools are valuable and which are troubling, students in nontechnical fields need to understand what an algorithm is and how one can be ethically deployed.

This project examines how Generative AI tools impact developers' coding performance, using empirical analysis and proprietary data from an IT organization.

Generative AI (GenAI) tools, such as GitHub Copilot, have emerged as transformative technologies in software development, offering the potential to enhance coding efficiency through real-time suggestions. However, the impact of GenAI on developers' coding performance remains underexplored. This study investigates the effects of GenAI usage on coding productivity by leveraging a difference-in-differences (DID) analysis of 27 weeks of proprietary data from a mid-sized global information technology organization.

DoppelgAIngers displays a false "mirror" image that looks eerily like you: dressed similarly, posed similarly, but who does not exist.

"Deep fakes" are images or videos that have been manipulated with AI to misrepresent the truth. These images spread virally online, giving credence to conspiracy theories and stoking political, social, and cultural tensions. These images force us to question the nature of visual truth. How can we investigate what we know to be true? How can we reconcile the reality of the physical world with the subjective nature of images? How can we recognize manipulated images when we see them? DoppelgAIngers is a web-based interactive piece and series of images.

This random forest algorithm categorizes old stars as exoplanet hosts based on transits found in Transiting Exoplanet Survey Satellite photometry.

Most stars have planets orbiting them, and some cross in front of the host star causing a temporary decrease in brightness known as a transit. These transits have measurable depth, period and duration. Only ~6% of known transiting planets orbit older host stars, and the rest orbit stars with ages similar to our sun1. As a result of stellar evolution, transits of old stars are sometimes missed by transit searches of young stars. The goal of this large-scale search is to quickly discover enough exoplanet candidates orbiting old stars to assist population level analysis.

Structure for leading an in-class creative activity, based in iterative theatrical devising and driven by group collaboration with generative AI.

This Lehigh AI Project provides a structure for a creative activity in order to experiment with and reflect upon the potential of AI for artistic collaboration. Over the course of a 75-minute in-class exercise, students (or other participants) engage in an iterative artistic process with text- and image-generating AI applications. Over a sequence of steps, students develop a creative work that incorporates poetic text, digital imagery, and embodied performance. The process combines individual effort and group decision-making, entangling both in artistic collaboration with AI.

We plan to focus on AI training to improve students' performance on the Oral Proficiency Interview (OPI). We will train the AI on the structure of the OPI using ACTFL proficiency guidelines and design prompts that simulate a natural, contextual target language environment.

The existing CFLingo platform supports conversations between Chinese as a Foreign Language (CFL) students and OpenAIs ChatGPT LLM. The platform facilitates a variety of exercises, including debates, roleplays, sentence creation, essay outlining, and essay drafting. We are currently in the process of augmenting the scripts supporting our existing CFLingo platform to use audio as both an input and output as well as text. The existing platform uses OpenAIs GPT-4-Omni LLM, which only uses text for input and output.

Developing scalable, interpretable AI tools integrating generative models to revolutionize computational materials discovery and enhance STEM education.

This project addresses challenges in computational materials modeling by creating scalable, interpretable AI tools. Current machine learning (ML) models face high computational costs, difficulty handling complex crystal symmetries, and a "black box" nature that limits interpretability. To address these issues, we aim to integrate nonnegative matrix factorization (NMF) with generative AI. NMF offers structure-aware capabilities, reducing computational costs and incorporating material symmetries, while generative AI produces high-fidelity synthetic datasets to augment limited data.

We are developing physics-guided machine learning techniques for cardiovascular systems, to obtain a personalized digital twin model for heart diseases.

Heart disease, characterized by changes in vascular, valvular, and ventricular systems, has been the leading cause of death in US since 1950. These diseases often happen acutely, making early detection essential. This interdisciplinary project aims to transform the clinical approach to early stage heart diseases, which is still dominant by traditional methods, through an AI-based approach. Our goal is to integrate data from wearable devices, mechanical measurements, and physics-based modeling, towards a personalized digital twin model for heart diseases.

This project explores how pre-professional journalists integrate AI into workflows, assessing its impact on ethics, creativity, and productivity.

This research examines the integration of artificial intelligence (AI) into the workflows of pre-professional journalists and explores its implications for journalistic ethics. With a focus on human-AI employment, the study seeks to understand how AI tools are perceived, adopted, and used by student journalists. Participants from a collegiate student newspaper will engage in an survey to first better understand the newsroom attitudes and culture around AI knowledge and engagement.