Research

College of IST awards seed grants to six research projects

Westgate Building at University Park is home to the Penn State College of Information Sciences and Technology.  Credit: Penn State. Creative Commons

UNIVERSITY PARK, Pa. — The Penn State College of Information Sciences and Technology (IST) recently selected six projects to receive funding from the college’s seed grant program.

The program provides funding to support research activities and generate preliminary results that will eventually lead to bigger projects involving external funding. Members of the IST faculty wrote short proposals that were reviewed by their peers. The highest-rated proposals were recommended to Andrea Tapia, dean of the College of IST, and awarded on a competitive basis.

“These projects demonstrate the interdisciplinary research that is a hallmark of the College of IST,” Tapia said. “They explore important societal challenges that can be addressed through information and technology and show tremendous potential for future success.”

The six projects selected for funding are

“Artificial Intelligence (AI) Meets Animal Science: Estimating and Forecasting Body Weights in Poultry Species” by Dongwon Lee, professor in the College of IST, along with John Boney, associate professor of poultry science, and Enrico Casella, assistant professor of data science for animal systems and Institute for Computational and Data Sciences co-hire, both in the Penn State College of Agricultural Sciences. In this project, the team aims to develop and evaluate, using AI tools, a cost-effective, automated, weight estimation system that improves poultry farm efficiency, reduces manual labor and enhances data-driven decision-making in animal production. By using several forms of AI, this study seeks to produce preliminary results and a prototype solution to accurately estimate body weights of poultry species using camera-taken depth images and to forecast future body weights based on time-series observations. The goal is to bridge the gap between experimental research and real-world farm applications, contributing to a more sustainable and efficient poultry industry.

“Computer Vision Meets Art History: New Perspectives on Women Artists of the Impressionist Movement” by James Z. Wang, distinguished professor in the College of IST, and Kathryn Brown from Loughborough University: Computer vision has transformed the study of visual art, offering new interpretive tools for historians and conservators. Yet, much of this work overlooks the contributions of female artists. This project will examine the stylistic innovations of Mary Cassatt and Berthe Morisot, who challenged gender norms, shaped Impressionism and offered fresh perspectives on 19th-century women’s experiences. Merging humanities and STEM disciplines, this project not only advances the study of underrepresented women artists in the Impressionist movement but also serves as a model for applying AI to complex and abstract problems in the humanities, bridging disciplines and fostering critical discussions on gender representation, diversity and the role of AI in cultural heritage.

“Generative AI Models as Social or Economic Agents” by Hadi Hosseini, associate professor in the College of IST, and Ran Shorrer, associate professor of economics in the Penn State College of the Liberal Arts: The rapid growth in the adoption of AI systems has transformed the ways social, political and commercial decisions are governed. Ensuring societal properties such as fairness within AI systems is essential not only for establishing procedural justice but also for enhancing trust, transparency and accountability in automated decision-making. This project aims to develop a framework for aligning AI models with human and social preferences by combining algorithmic techniques from computer science with axiomatic methods rooted in economic theory and social sciences. In particular, the objectives include evaluating generative AI models for alignment with human preferences, analyzing the economic and algorithmic reasoning embedded in these models and developing techniques to fine-tuning and better alignment of AI solutions with societal and economic preferences.

“Large Language Models (LLMs) and the future of work in Software and Information Technology (IT)” by Amulya Yadav and Kenneth Huang, both associate professors in the College of IST: This project aims to measure and analyze the extent to which the roles of software developers and IT administrators are at risk of being automated or augmented by LLMs. The researchers will identify day-to-day tasks in a worker’s job that can be completely taken over by LLMs, which tasks can be completed with the help of worker/LLM teaming and which tasks are still beyond the scope of LLMs. With these insights, the researchers want to understand how required skills sets for software developers an IT administrators may evolve in response to LLM capabilities. This includes identifying skills that may increase in demand because of tasks that LLMs are not adept at performing, while also identifying skills which will decrease in demand because LLMs may be able to do some tasks proficiently.

“Mitigating Hallucinations in LLM Generation via Representation Flow Correction” by Jinghui Chen, assistant professor in the College of IST: LLMs have demonstrated remarkable performance across various natural language processing (NLP) tasks. However, they are prone to hallucination — a phenomenon where the generated content appears plausible but is ultimately misleading or inconsistent with established knowledge — which significantly undermines the trustworthiness of LLMs. The teams' research aims to design a novel framework to better mitigate hallucinations and elicit truthful generations from LLMs to ensure their reliable deployment in practical applications. Addressing these critical issues contributes to the development of more reliable and responsible LLM systems. This work may also inspire researchers in the broader LLM community, fostering advancements in more reliable and trustworthy LLMs. Furthermore, the teams' methods may also significantly boost the potential of adopting LLMs in critical application domains such as medicine, health care, law and finance.

“Toward Deeper Cognitive Engagement: Mitigating Cognitive Offloading through Adaptive AI-Assisted Pretesting” by Mahir Akgun, assistant teaching professor, and Peng Liu, Raymond G. Tronzo, MD Professor of Cybersecurity, both in the College of IST: In STEM education, students often struggle to engage deeply with new material and retain key concepts, particularly in foundational courses where diverse prior knowledge can create significant challenges. Pretesting, a pedagogical strategy that involves asking learners questions about a topic before they formally learn it, has been shown to prime memory, activate prior knowledge and direct attention to critical information. The teams' preliminary research extended this approach by integrating AI-assisted pretesting into an undergraduate statistics course, demonstrating significant improvements in student performance but with two critical challenges: instructor burden and the evolving nature of cognitive offloading. This proposal aims to develop and pilot an adaptive AI-assisted pretesting system and empirically investigate its role in a human–AI partnership context. By focusing on adaptive pretesting and its connections to cognitive offloading in an undergraduate statistics course — and ultimately other STEM courses — the team aims to generate actionable insights into designing scalable, effective teaching strategies for the generative AI era.

Last Updated April 25, 2025

Contact