June 2, 2025
An interdisciplinary perspective – one that combines humanities with technical knowledge – is essential for building technology that is ethical, inclusive, and effective in real-world contexts. With events like the Mid-Atlantic Student Colloquium on Speech, Language, and Learning (MASC-SLL), researchers at the School of Computing and Information (SCI) have the opportunity to broaden their outlooks.
“About 10 students came to me, introduced their research, and showed interest in my research after the talk. I’m especially glad to see that so many different people are interested in this issue – I think it’s a really important problem to work on,” reflected Lorraine Li, an Assistant Professor at SCI and a keynote speaker at the April 5 colloquium.
Li presented research exploring the ways in which cultural contexts are often under considered in Large Language Model (LLM) outputs, therefore decreasing the accuracy and integrity of the technology. For example, an LLM may associate boiling water with cooking or making tea – however, it may fail to consider that in regions with limited water resources, it would be more accurate for an LLM to associate boiling water with germ elimination.
The MASC-SLL colloquium facilitates discourse between professionals in the field that can help to address cultural and social issues in developing technologies.
Zhexiong Liu, a PhD student in computer science, presented research that bridges education and computing. His team developed eRevise+RF, a system piloted in select schools across Pennsylvania and Louisiana, which supports students in grades 4 through 8 as they write and revise argumentative essays.
“I need to learn from experts in education to understand how their ideas can be integrated into the project using computer science techniques. I’ve found collaborating with researchers from diverse backgrounds is essential for tackling the complexities of multidisciplinary work,” Liu said.
Another one of SCI’s PhD students in computer science, Bhiman Kumar Baghel, presented research about making AI language models more accurate by manually editing how they store and process information. He identified common issues in existing editing methods and addressed them by gaining a deep theoretical understanding of these model’s underlying assumptions, limitations, and behavior in different settings, then developed new solutions based on that knowledge.
“Computing is crucial because it allows us to teach machines to understand and communicate in human language. As language models take over routine tasks, people can focus more on creativity and innovation, ultimately improving how we learn, connect, and solve problems as a society. But to make these systems trustworthy, we need to understand how they work,” said Baghel.
One of the essential principles of ethical technology is being sure that it is transparent and understandable. Baghel’s work helps to build a relationship of trust between humans and tech, therefore helping to build a more ethical digital future.
The conference created a space for SCI students and staff to develop their skills in presenting, networking, and collaborating to help further innovation in the field.
“The symposium was a great opportunity to meet students from other universities and get a sense of the current research directions in NLP,” said Andrew Aquilina, another PhD student at SCI.