2025-05-12: ACM Capital Region Celebration of Women in Computing (CAPWIC) 2025 Trip Report

    

The ACM Capital Region Celebration of Women in Computing (CAPWIC) 2025 conference took place on March 29, 2025, and it was hosted by the George Washington University in Washington, D.C. CAPWIC provides a low-cost, regionally tailored platform for women and minorities in computing, including students, faculty, professionals, and supporters. 

This year, CAPWIC welcomed 193 attendees from universities, high schools, companies, and non-profit organizations. The majority were undergraduate students (41%), followed by graduate students (36%), college and university faculty (15%), industry professionals (4%), and high school students (3%). Eight Ph.D. students in Computer Science from ODU attended the CAPWIC 2025 conference. Among them, five students from ODU's  Web Science and Digital Libraries (WS-DL) research group participated in person, presenting research shorts and posters.


Ph.D. students from ODU Computer Science at the CAPWIC 2025 conference at George Washington University

Opening Remarks and Keynote #1

The CAPWIC 2025 conference started with opening remarks by the conference chair, Dr. Megan Olsen from Loyola University Maryland


The opening remarks by the CAPWIC 2025 conference chair, Dr. Megan Olsen


Dr. Quincy K. Brown delivered the first keynote, “Oh, The Places You'll Go!” Dr. Brown was most recently the Director of Space STEM and Workforce Policy at the National Space Council. She is a computer scientist, former assistant professor at Bowie State University, and former Senior Policy Advisor in the White House Office of Science and Technology Policy


Dr. Brown shared how her background in computer science has informed her work across research, education, and policy, emphasizing the importance of innovation through collaboration and interdisciplinary engagement. She highlighted three key principles, namely policy, practice, and people, as the main operating principles in the working environment. Her keynote concluded with a powerful reflection: "Whose life do you want to be better because you are here?"




Dr. Quincy K. Brown gives the morning CAPWIC 2025 keynote

Parallel Sessions - Cybersecurity

Following the morning keynote, two parallel sessions were held: Cybersecurity and AI. In the Cybersecurity session, ODU's own CS PhD student and senior lecturer, Susan Zehra, presented “Mitigating Cyber Threats in V2V and V2I) Networks: A Security-Centric Approach”. Her research focuses on Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication, which involves vehicles interacting with each other and with traffic control systems. Her method combines dynamic key management with RSUs (Roadside Units), public-key cryptography, and blockchain-based fallback mechanisms to counter cyber threats. As a result, her approach significantly reduced the breach rate for key exchange between vehicles while ensuring reliability and user anonymity.



The other two presentations in the session were: “Combining Open-Source Intelligence (OSINT) with AI for Threat Detection” by Jackline Fahmy (Marymount University) and “A Hierarchical Deep Reinforcement Learning Chatbot for Cybergrooming Prevention” by Heajun An (Trustworthy Cyberspace Lab - tClab at Virginia Tech). Fahmy's study explored how AI and NLP techniques can enhance threat detection across social media platforms and the Dark Web. Heajun An's research introduces an AI chatbot that adapts to different vulnerability levels in teenagers by using dynamic interactions to prevent cyber-grooming.

Student Research Posters

Kritika Garg from WS-DL presented her poster “Redirects Unraveled: From Lost Links to Rickrolls.” The research examined 11 million redirecting URIs to uncover patterns in web redirections and their implications on user experience and web performance. While half of these redirections successfully reached their intended targets, the other half led to various errors or inefficiencies, including some that exceeded recommended hop limits. Notably, the study revealed "sink" URIs, where multiple redirections converge, sometimes used for playful purposes such as Rickrolling. Additionally, it highlighted issues like "soft 404" error pages, causing unnecessary resource consumption. The research provides valuable insights for web developers and archivists aiming to optimize website efficiency and preserve long-term content accessibility.




Lesley from WS-DL presented her work, “US Computer Science Faculty Pipeline Dashboard.” This dashboard was a project made in CS 725 Information Visualization in spring 2022 taught by Dr. Michele C. Weigle. The dashboard aligned Clauset’s “Systematic inequality and hierarchy in faculty hiring networks” dataset with IPEDS data to visualize the pipeline of female CS PhD graduates to CS tenure track faculty positions at US institutions. The dashboard can be used to compare the graduate student gender diversity between different universities, examine historical trends in female PhD CS programs, and examine trends in regional CS hiring.




Thu Nguyen from the Bioinformatics and Parallel Computing group at ODU presented “Enhancement of Deep Learning for Segmentation of Protein Secondary Structures from Cryo-EM,” at the poster session. Their work enhances the efficiency of DeepSSETracer, a deep learning framework for cryo-electron microscopy segmentation, by partitioning large cryo-EM maps and merging the outputs. This approach reduces memory usage and processing time, enabling smoother handling of large datasets within the ChimeraX visualization tool.



Parallel Sessions - CS Education

The session on CS education was chaired by Briana Morrison and featured data science research on admissions biases and success rates in CS courses.


Anaya Prakash presents at CAPWIC 2025


Anaya Prakash of Virginia Tech presented “Who Gets In? The Role of AI in Shaping the Next Generation of Computer Scientists.” She analyzed a closed dataset of a large public university’s MS CS admissions. She found patterns in the admissions data, including a preference for men from either the US or China being 2.5X more likely than other subgroups. She also found that attributes like age, which were not intended to be a factor in admissions, were key predictors in models such as Random Forest.


Nawar Wali presents at CAPWIC 2025


Next, Nawar Wali of Virginia Tech presented her work, “Machine Learning Insights into Academic Success in CS3: The Role of Mathematics and CS Coursework.” She analyzed how mathematics coursework is correlated with success in Data Structures and Algorithms (CS3). She analyzed 10 years of Virginia Tech CS3 data, including CS and math courses with associated grades for 3900 students. She found that discrete structures and computer organization were the highest correlated courses with CS3 success. She also found that students who pass linear algebra earlier also have a higher success rate in CS3.

Parallel Sessions - Artificial Intelligence

The Artificial Intelligence session was hosted by Robbie Hott from the University of Virginia.

Victoria Wiegand from Villanova University presented “Data Collection Pipeline to Diversify AI Training Data.”  Their work addresses the cultural biases present in AI vision-language models, which are often trained on predominantly Western data sources. As a result, these models perform poorly when interpreting or generating content from non-Western contexts, frequently misrepresenting communities with inaccurate or stereotypical imagery. To help bridge this digital divide, the researchers developed a low-cost, community-driven data collection pipeline. Partnering with a university service trip, they trained participants in ethical data collection and gathered images via a WhatsApp-linked web form. These images will be compiled into a publicly available dataset aimed at helping developers diversify model training data and improve cultural representation.


Victoria Wiegand presents at CAPWIC 2025


Hajra Klair from Virginia Tech presented “Agentic AI for the Rescue: Factual Summarisation of Crisis-Related Documents.” Their work presents a new approach to summarizing crisis-related documents using large language models. Evaluated on the CRISISFacts dataset covering 18 real-world events, the two-phase architecture first retrieves documents based on entity prominence and then generates summaries guided by structured, crisis-specific queries. This approach reduces hallucinations and enhances information coverage, aligning the summaries with the needs of emergency response officials. 

Hajra Klair presents at CAPWIC 2025


Yeana Lee Bond from Virginia Tech presented “Driver Facial Expression Classification: A Comparative Study of Computer Vision Techniques.”  Their work explores how well different machine learning models can detect driver emotions—happy, angry, and neutral—using facial expressions. They evaluated EfficientNet, Vision Transformer, and CNN-based models on a driver emotion dataset. EfficientNet stood out for its speed and high accuracy, while the study also highlighted how model design and data quality play a bigger role than just model size.

Parallel Sessions -  Games and Virtual Reality

The Games and Virtual Reality session was chaired by Dr. Jin-Hee Cho, Associate Professor and Director of the Trustworthy Cyberspace Lab (tClab) in the Department of Computer Science at Virginia Tech. The session featured three short research presentations.


Yasasi from NIRDS Lab and WS-DL presented “Examining Visual Attention in Gaze-Driven VR Learning: An Eye-Tracking Study” at the Games and Virtual Reality session. Their work presents a framework for analyzing visual attention in a gaze-driven VR learning environment using a consumer-grade Meta Quest Pro VR headset with a built-in eye-tracker. Yasasi discussed how their study contributes by proposing a novel approach for integrating advanced eye-tracking technology into VR learning environments, specifically utilizing consumer-grade head-mounted displays.



Parallel Sessions - Social Media

The session on social media was hosted by Zhoujun Duan from James Madison University.


Sanjana Kumari from Virginia Tech presented “Evaluating Children's Ability to Distinguish Between Traditional and AI-Generated Media”. Their study investigates whether children can distinguish between human-authored and AI-generated content through a structured intervention comprising surveys and an educational workshop. The authors note that while adults are increasingly adapting to tools like ChatGPT, we still lack a clear understanding of how children perceive and process these technologies.


Rebecca Ansell from Georgetown University presented “Assessing Public Perception of AI-Generated Social Media Content of the 2024 U.S. Presidential Debate”. Their study explored key research questions such as: Can humans distinguish between AI-generated and human-authored content on social media? And what characteristics make a social media post appear human? To investigate this, they collected a dataset of X posts and YouTube comments related to the 2024 presidential debate, and supplemented it with content generated using ChatGPT. Human annotators were then employed to label whether each piece of content was AI-generated or not. The researchers used the Bradley-Terry model to analyze the ratings and evaluate human perception patterns. The findings showed that annotators could generally differentiate between AI and human content. Interestingly, the study also examined how sentiment influenced perceived humanness, positively toned posts were more likely to be perceived as AI-generated, while negative or offensive content was often seen as human-authored. The results highlight the role of tone, civility, and emotional cues in shaping public perception.

Himarsha from WS-DL presented “Infrastructure for Tracking Information Flow from Social Media to U.S. TV News”. Their work focused on understanding how social media content is amplified through mainstream media by expanding its reach to new audiences. They are using the data from Internet Archive’s TV News Archive to explore how social media content flows into TV news and the contexts in which it is incorporated.



Parallel Sessions - Human-Computer Interaction

In parallel to the Social Media session, the Human-Computer Interaction (HCI) session was held, featuring three short research presentations. It was chaired by Dr. Jin-Hee Cho from Virginia Tech.


The session began with Anika Islam from George Mason University, presenting their research titled “Leveraging Smartwatch Sensors for Detecting Off-Task Behaviors of Neurodivergent Individuals”. The study aims to improve workplace success for neurodivergent individuals by using smartwatch sensor data to identify off-task behaviors and deliver personalized interventions. In the current phase, data were collected from 25 neurodivergent young adults engaged in a manual task within a controlled lab environment. Anika outlined their future plans to apply machine learning techniques to analyze the data, with the goal of developing real-time, tailored interventions that improve productivity in the workplace.


The next presenter, Marissa Hirakawa, an undergraduate Computer Science student from Virginia Tech, presented “Usability Heuristics and Large Language Models: Enhancing University Website Evaluations.” Their study focused on exploring the use of large language models (LLMs) to support usability evaluations of university websites, based on Nielsen’s 10 usability heuristics. Their findings show that while LLMs can help uncover usability issues often missed in manual reviews, human verification remains essential as LLMs occasionally hallucinate usability issues. Their future directions include refining LLM evaluation processes and incorporating multimodal data such as user action logs and screenshots to enhance assessment accuracy.


Kumushini from NIRDS Lab and WS-DL presented “Advanced Gaze Measures for Analyzing Joint Visual Attention.” This research explored how user pairs coordinate their joint visual attention (JVA) by using egocentric and eye-tracking data. In their user study, participants engaged in a collaborative screen-based visual search task while wearing Project Aria smart glasses. Their findings suggest that users who maintained similar attention behaviors (ambient/focal) over time exhibited more frequent and sustained moments of joint attention compared to those with differing attention behaviors. In future work, they plan to further refine the methodology by integrating machine learning techniques to automatically identify and classify different patterns of ambient and focal visual attention during collaborative tasks.




Parallel Sessions - Machine Learning


ODU’s Eleni Adam presents at CAPWIC 2025


In the machine learning session, Eleni Adam from the Bioinformatics Lab at ODU presented “Analysis of Subtelomere and Telomere Regions of Cancer Genomes on the Cloud.” Eleni examined subtelomeres in cancer patients. She used ODU’s Wahab Cluster to carry out her research. In this work, she implemented a computational pipeline to enable the subtelomere analysis. In reality, each patient’s DNA takes hours to run. Her ultimate goal is to understand cancer and the subtelomere’s role in it. Her work is available at https://github.com/eleniadam/storm.

Keynote #2, Awards, and Closing Remarks 


Becky Robertson gave the closing keynote at CAPWIC 2025


Becky Robertson, Vice President at Booz Allen Hamilton delivered the closing keynote. Her talk centered around the concept of inspiration. She engaged with the audience by asking about what or who inspires us personally. She encouraged us to channel that inspiration into meaningful actions, overcome challenges, and pursue our goals. 


Following the keynote, it was time for the award ceremony and the closing remarks. Several awards were presented, including the Best Research Short Award, Honorable Mention Research Short Awards, Flash Talk Awards, and Best Poster Awards, which recognized outstanding contributions from both graduate and undergraduate participants. 


It was especially exciting to see Thu Nguyen from the Bioinformatics and Parallel Computing Group at ODU winning the Best Poster Award in the graduate category.




Thu Nguyen from the Bioinformatics and Parallel Computing Group at ODU received the Best Poster Award in the Graduate Category


During the closing remarks session, the next year’s organizing committee announced that CAPWIC 2026 will be at Virginia Tech’s Innovation Campus in Alexandria, VA. 

Wrap-up

For all of us, it was our first time attending the CAPWIC conference in person. CAPWIC 2025 provided an inspiring platform to exchange new ideas and showcase innovative research within the tech community, encouraging greater participation among women and minorities in computing. The CAPWIC 2025 conference was held in Washington, D.C., during the peak of the cherry blossoms. We had the opportunity to take part in the National Cherry Blossom Festival at the Tidal Basin, enjoying the beautiful sight of the city covered in pink and white blossoms. It was a memorable experience to see the capital come to life with the colors of spring.

Ph.D. students from WS-DL at Tidal Basin, Washington, D.C.



--Yasasi (@Yasasi_Abey), Lesley (@lesley_elis), Himarsha (@HimarshaJ), and Kritika (@kritika_garg)


Comments