Current Research

Beyond the Individual: Community-Engaged Design and Implementation of a Framework for Ethical Online Community Research

Advised by Lana Yarosh, Stevie Chancellor, Joe Konstan, and Loren Terveen

Community-level harms and benefits are frequently not captured by ethics reviews focusing on individual-level harms. Moreover, the research guidelines that do exist may be significantly different from the expectations of members of online communities. To address this gap, we are developing and validating a framework for ethical conduct of research with online communities through participatory and community-engaged approaches of several critical online communities.

Civil vs. Global: Explore and Depolarize U.S. Reader News Consumption With Personalized LLM Locality Context

Co-first Author Ruixuan Sun | Advised by Joe Konstan

Diverse news consuption is important facet of social reslience. Increasing the diversity of news recommendation has deminishing returns on consuption diversity. To strength the influence of diversity calibration, we are using large language models to highlight personalized events embedded within diverse news on the news recommendation platform POPROX.

Surrogate Anonymization of Personally Identifiable Information in Math Tutoring Dialogues: A Dataset

Eedi | Advised by Simon Woodhead

Large edutech datasets are highly saught after in learning engineering. However, online dialogues between tutors and student frequently contain much perosonally identifiable information (PII). In partnership with Eedi, an online math learning platform, we design PIIvot, an ML and LLM backed hidden-in-plain-sight approach to PII obfuscation in tutor-student dialogues, to faciliate sharing of the largest tutoring dataset of its kind.

Anonymity, Verbal Person-Centeredness, and Identity in Online Recovery and Social Support

Advised by Lana Yarosh

Social support research has investigated reciprocal self-disclosure and when it occurs in practice, but little is known about the effects of reciprocal self-disclosure on support seekers. Inspired by prior work investigating the effects of verbal person centeredness (VPC), we explore the interaction effects of VPC and identity in 12-step recovery, a context that promotes anonymity.

Experience

Research Assistant

August 2021 - (Current)

GroupLens

Investigating the design of social computing systems through empirically studying online communities; using mixed methods approaches to design, implement, and evaluate human centered systems (see Publications).

Machine Learning Research Consultant

November 2024 - (Current)

Eedi

Using PIIvot to obfuscate PII in Eedi datasets; evaluating potential risks to platform users and dataset quality to ensure responsible and effective data usage in downstream tasks.

Machine Learning Research Intern

May 2024 - August 2024

Eedi

Led the engineering of the open source package PIIvot; fine-tuned Deberta-based NER models to achieve 98% balanced test accuracy; implemented feedback-based repompting for LLM-generated surrogates for PII obfuscation.

Software Engineer

July 2019 - May 2021

Microsoft

Designed and implemented services at scale to host Dynamics 365 apps in Docker containers on Service Fabric clusters; led the design and implementation of my team's work on Project Uno (the seamless integration of CDS and Dynamics 365 infrastructure systems)

Co-Founder/Partner

September 2020 - July 2021

Fivestack

Created software solutions for joint venture partnerships; owned the full-stack implementation of the iOS/Android application for the successful Kickstarter doghouse

IS Software Developer Intern

May 2018 - August 2018

Federated Mutual Insurance

Developed SQL for companywide Multi Product Data Mart reports; automated the company's process for building SSRS reports with VBA Script; presented R&D for an Agile approach to software development to management.

Research Intern

May 2017 - August 2017

UNC Charlotte HCI Lab

Worked independently on the backend recommender system by using machine learning to classify recipe data; conducted empirical study to develop personalized model for curiosity (see Awards).

Publications

Beyond the Individual: A Community-Engaged Framework for Ethical Online Community Research

Matthew Zent, Seraphina Yong, Dhruv Bala, Stevie Chancellor, Joe Konstan, Loren Terveen, Stetlana Yarosh

In-Review at CSCW

Anonymity in Online Recovery: Measuring the Effects of Verbal-Person Centeredness and Identity Disclosures in Recovery Support

Matthew Zent, Kathleen Shae, Svetlana Yarosh

In-review at CSCW

Peer Recommendation Interventions for Health-related Social Support: a Feasibility Assessment

Zachary Levonian, Matthew Zent, Ngan Nguyen, Matthew McNamara, Loren Terveen, Svetlana Yarosh

Accepted to CSCW 2025

Awards

ACM CHI Graduate Student Research Competition

1st Place - 2023

Anonymous Online Support: Investigations of Identity and Heterogeneous Groups in Online Recovery Support

Extended Abstract | CHI 2023 Winners

Unstoppable Together Hackathon

1st Place - 2020

24-hour hackathon innovating for Jira, Confluence, and Atlassian users; automation, integrations, machine learning, or Tempo themed; created MyTempo, a Alexa based skill integrating with Jira and Tempo.

Video | Github | Blog

Fargo Hackathon - Emerging Prairie

1st Place - 2019

24-hour hackathon for behavioral and mental health issues; developed My Hero, a mobile web app written using Blazor for connecting individuals with peer-support mentors.

Github | Blog

Microsoft Fargo Hackathon

1st Place People's Choice - 2019

72-hour open themed hackathon; created Cyber Attack, an educational game developed in Unity for use in teaching important cyber security concepts to non-technical individuals.

Github

UNC Charlotte Research Symposium

1st Place - 2017

Poster talk and research symposium for REU students working in HCI, Visualization, and Vision and Image Analysis Labs; presented personal contributions to Q-Chef: A User Experience Based on AI Models of Curiosity

Poster | Abstract | Blog